Quality, Trade, and Exchange Rate Pass-Through¤y

Natalie Chen

University of Warwick, CAGE,

CESifo and CEPR

Luciana Juvenal

International Monetary Fund

February 26, 2015

Abstract

We investigate theoretically and empirically the e¤ects of real exchange rate changes on the

behavior of …rms exporting multiple products with heterogeneous levels of quality. Our model

predicts more pricing-to-market and a smaller response of export volumes to a real depreciation for

higher quality products. We …nd strong support for the model predictions using a unique data set

of Argentinean …rm-level wine export values and volumes between 2002-2009 combined with experts

wine ratings to measure quality. Our results have implications for the choice of optimal monetary

and exchange rate policies, and for the understanding of international market segmentation.

JEL Classi…cation: F12, F14, F31

Keywords: Exchange rate pass-through, pricing-to-market, quality, unit values, exports, multi-

product …rms, wine.

¤We thank the Centre for Competitive Advantage in the Global Economy (CAGE) at the University of Warwick, theFederal Reserve Bank of St Louis, and the International Monetary Fund for …nancial support. For helpful comments,we thank Nicolas Berman, Andrew Bernard, Illenin Kondo, Logan Lewis, Philippe Martin, Guillermo Noguera, DennisNovy, Lucio Sarno, Cédric Tille, Olga Timoshenko, Zhihong Yu, our formal discussants Raphael Auer, Julian di Giovanni,Ferdinando Monte, Tim Schmidt-Eisenlohr, and Charles van Marrewijk, and participants at the CAGE Research Meetings2013, CEPR ESSIM 2014, CEPR Exchange Rates and External Adjustment Conference Zurich 2014, EITI Phuket 2014,ETSG Birmingham 2013, Federal Reserve Bank of Atlanta and NYU’s Stern School of Business 2nd Annual Workshopon International Economics, INFER Conference on Regions, Firms, and FDI Ghent 2014, Midwest International TradeMeetings 2013, Tri-Continental International Trade Policy Conference Beijing 2014, DC Trade Study Group, GenevaTrade and Development Workshop, Inter-American Development Bank, International Monetary Fund, London Schoolof Economics, Lund, Nottingham, Utrecht School of Economics, and Warwick. We are grateful to Linda Goldberg forsharing the data on distribution costs. Brett Fawley provided excellent research assistance. The views expressed are thoseof the authors and do not necessarily re‡ect o¢cial positions of the International Monetary Fund.

yNatalie Chen, Department of Economics, University of Warwick, Coventry CV4 7AL, UK. E-mail:N.A.Chen@warwick.ac.uk (corresponding author) and Luciana Juvenal, International Monetary Fund. E-mail:ljuvenal@imf.org.

1 Introduction

Exchange rate ‡uctuations have small e¤ects on the prices of internationally traded goods. Indeed,

empirical research typically …nds that the pass-through of exchange rate changes to domestic prices is

incomplete, leading to deviations from the Law of One Price.1 Possible explanations for partial pass-

through include short run nominal rigidities combined with pricing in the currency of the destination

market (Engel, 2003; Gopinath and Itskhoki, 2010; Gopinath, Itskhoki, and Rigobon, 2010; Gopinath

and Rigobon, 2008), pricing-to-market strategies whereby exporting …rms di¤erentially adjust their

markups across destinations depending on exchange rate changes (Atkeson and Burstein, 2008; Knetter,

1989, 1993), or the presence of local distribution costs in the importing economy (Burstein, Neves, and

Rebelo, 2003; Corsetti and Dedola, 2005).2

Thanks to the increasing availability of highly disaggregated …rm- and product-level trade data,

a strand of the literature has started to investigate the heterogeneous pricing response of exporters

to exchange rate changes.3 Amiti, Itskhoki, and Konings (2014) …nd that Belgian exporters with

high import shares and high export market shares have a lower exchange rate pass-through. Berman,

Martin, and Mayer (2012) show that highly productive French exporters change signi…cantly more their

export prices in response to real exchange rate changes, leading to lower pass-through. Chatterjee, Dix-

Carneiro, and Vichyanond (2013) focus on multi-product Brazilian exporters and …nd that within …rms,

pricing-to-market is stronger for the products the …rm is most e¢cient at producing.

Evidence on the role of product-level characteristics in explaining heterogeneous pass-through re-

mains scarce, however. In this paper, we explore how the heterogeneous pricing-to-market behavior

of exporters, which leads to incomplete pass-through, can be explained by the quality of the goods

exported. We model theoretically the e¤ects of real exchange rate changes on the pricing decisions of

multi-product …rms that are heterogeneous in the quality of the goods they export, and empirically

investigate how such heterogeneity impacts exchange rate pass-through for export prices. Assessing the

role of quality in explaining pass-through is a challenge as quality is generally unobserved. To address

this issue we focus on the wine industry, which is an agriculture-based manufacturing sector producing

di¤erentiated products, and combine a unique data set of Argentinean …rm-level destination-speci…c

export values and volumes of highly disaggregated wine products with experts wine ratings as a directly

observable measure of quality.4

The …rst contribution of the paper is to develop a theoretical model to guide our empirical spec-

i…cations. Building on Berman et al. (2012) and Chatterjee et al. (2013), we extend the model of

Corsetti and Dedola (2005) by allowing …rms to export multiple products with heterogeneous levels

of quality. In the presence of additive (per unit) local distribution costs paid in the currency of the

1For a survey of the literature, see Burstein and Gopinath (2014) and Goldberg and Knetter (1997).2Also, Alessandria and Kaboski (2011) stress the role of search frictions as a source of incomplete pass-through.

Nakamura and Steinsson (2012) argue that price rigidity and product replacements lead aggregate import and exportprice indices to appear smoother than they actually are, biasing exchange rate pass-through estimates.

3Many papers examine the response of import prices (which include transportation costs) or consumer prices (whichfurther include distribution and retail costs) to changes in currency values (e.g., Campa and Goldberg, 2005, 2010;Gopinath and Rigobon, 2008). For earlier evidence from the perspective of exporters, see Goldberg and Knetter (1997),Kasa (1992), and Knetter (1989, 1993). For more recent evidence at the …rm-level, see Campos (2010), Fitzgerald andHaller (2014), Fosse (2012), Li, Ma, Xu, and Xiong (2012), and Strasser (2013).

4Other papers that focus on speci…c industries include Auer, Chaney, and Sauré (2014) and Goldberg and Verboven(2001) for cars, Hellerstein (2008) for beer, and Nakamura and Zerom (2010) for co¤ee.

1

importing country, which are assumed to be higher for higher quality goods, the model shows that the

demand elasticity perceived by the …rms falls with a real depreciation and with quality. As a result, in

response to a change in the real exchange rate, exporters change their prices (in their own currency)

more, and their export volumes less, for higher quality products. Once we allow higher income coun-

tries to have a stronger preference for higher quality goods, as the evidence from the empirical trade

literature tends to suggest (e.g., Crinò and Epifani, 2012; Hallak, 2006), the heterogeneous response of

prices and quantities to exchange rate changes due to di¤erences in product quality is predicted to be

stronger for exports to higher income destination countries.

The second contribution of the paper is to bring the predictions of the model to the data. The

…rm-level trade data we rely on are from the Argentinean customs which provide, for each export ‡ow

between 2002 and 2009, the name of the exporting …rm, the country of destination, the date of the

shipment, the Free on Board (FOB) value of exports (in US dollars), and the volume (in liters) of

each wine exported. As we do not directly observe prices, we compute FOB export unit values as a

proxy for export prices at the …rm-product-destination level, using the data on the value and volume

exported. In order to assess the quality of wines, we rely on two well-known experts wine ratings, the

Wine Spectator and Robert Parker. Our approach to measuring quality is similar to Crozet, Head, and

Mayer (2012) who match French …rm-level exports data of Champagne with experts quality assessments

to investigate the relationship between quality and trade.

Our data are well-suited for identifying heterogeneous pass-through due to di¤erences in product

quality. First, the level of disaggregation of the customs data is unique because for each wine exported

we have its name, grape (Chardonnay, Malbec, etc.), type (white, red, or rosé), and vintage year

(i.e., the year in which the grapes are harvested). With such detailed information we can de…ne a

“product” in a much more precise way compared to the papers that rely on trade classi…cations such as

the Combined Nomenclature (CN) or the Harmonized System (HS) to identify traded products (e.g.,

Amiti et al., 2014; Berman et al., 2012; Chatterjee et al., 2013). Second, as the quality scores from both

the Wine Spectator and Parker rating systems are assigned to a wine according to its name, grape,

type, and vintage year, the trade and quality data sets can directly be merged with each other. As

a result, when using the Wine Spectator ratings that have the largest coverage of Argentinean wines,

the sample we use for the estimations includes 209 multi-product …rms exporting 6,720 di¤erent wines

with heterogeneous levels of quality (this contrasts with Crozet et al., 2012, who cannot distinguish

between the di¤erent varieties sold by each …rm, and therefore assume that each …rm exports one type

of Champagne only). Aggregating our data at the 12-digit HS level would reduce our sample size by a

factor of six, which would in turn signi…cantly lower the within-…rm variation in the quality of exported

wines as the 209 exporters would only be selling at most …ve di¤erent “products.” Third, the level of

disaggregation of our data ensures that compositional or quality changes do not a¤ect movements in

unit values. For instance, Feenstra and Romalis (2014) demonstrate that the variation in trade unit

values de…ned at the 4-digit SITC level is mostly attributable to di¤erences in product quality. Fourth,

thanks to the granularity of our data, our pass-through estimates are not a¤ected by the “product

replacement bias” put in evidence by Nakamura and Steinsson (2012), which may be a problem when

using aggregate prices data (also, see Gagnon, Mandel, and Vigfusson, 2014). Finally, export values

are FOB and therefore measure the revenue received by exporters at the border, excluding nontradable

components such as transportation costs, tari¤s, or distribution and retailing costs in the importing

country.

2

To the best of our knowledge, our paper is the …rst to exploit such highly disaggregated product-level

exports data to investigate empirically the pricing strategies of exporters in response to real exchange

rate ‡uctuations.5 Consistent with other …rm-level studies, we …nd that pass-through for export prices

is large: in our baseline regression, in response to a ten percent change in the real exchange rate,

exporters change their export prices (in Argentinean pesos) by 1.9 percent, therefore pricing-to-market

is low and pass-through is high at 81 percent. Higher quality wines are more expensive and are exported

at a higher price. Most interestingly, we show that the response of export prices to real exchange rate

changes increases with the quality of the wines exported, or in other words pricing-to-market rises with

quality (i.e., pass-through decreases with quality). This e¤ect is economically important as pricing-to-

market is found to be 36 percent higher for the top than for the lowest quality range. This heterogeneity

in the response of export prices to exchange rate changes remains robust to di¤erent measures of quality,

samples, and speci…cations. We also examine the heterogeneous response of export volumes to real

exchange rate ‡uctuations. Export volumes increase in response to a real depreciation, but by less for

higher quality goods. Finally, we …nd that the e¤ect of quality on the responses of export prices and

volumes to changes in the exchange rate is larger for exports to higher income countries. Overall, the

predictions of our model are strongly supported by our empirical results.

One concern with our estimations is the potential endogeneity of quality in explaining unit values

and export volumes. Although both the Wine Spectator and Parker rating systems are based on blind

tastings where the price of each wine is unknown, the tasters are told the region of origin or the

vintage year of each wine. As a result, we cannot rule out that the basic information that the experts

hold about each wine, in combination with their own subjective preferences, a¤ects in a way or the

other their scores, leading to an endogeneity bias. In addition, as wine producers are likely to set

their prices depending on the ratings that are given to their wines, measurement error in the quality

scores is another potential source of endogeneity. In order to overcome this issue, we use appropriate

instruments for quality based on geography and weather-related factors, including the total amount of

rainfall and the average temperatures during the growing season for each province where the grapes

are grown, as well as the altitude of each of the growing regions. Our …ndings remain robust to the

instrumentation of quality.

We also provide some extensions to our benchmark speci…cations. First, when allowing for non-

linearities in the e¤ects of quality, the elasticity of unit values to the exchange rate is shown to be

much larger for the top quality wines, a …nding which can be reconciled with our assumption that

higher quality goods have higher (per unit) distribution costs. Second, as predicted by the model,

pricing-to-market increases with local distribution costs in the importing economy. Third, in addition

to increasing with quality, pricing-to-market also increases with …rm-level productivity, as in Berman

et al. (2012) and Chatterjee et al. (2013). Fourth, once we divide the sample between high and low

quality exporters, high quality …rms price-to-market more than the others. Finally, as in Amiti et al.

(2014), …rms price-to-market more in the countries where they own a larger share of the market.

Although our analysis concentrates on a speci…c sector in a single country, evidence suggests that

cross-country and cross-industry di¤erences in the quality of traded goods are large (Feenstra and Ro-

5Papers using similarly detailed product-level data usually observe retail or wholesale prices rather than trade prices(e.g., Hellerstein, 2008; Nakamura and Zerom, 2010).

3

malis, 2014).6 Assuming that our results can be extended beyond the Argentinean wine industry, they

can therefore be expected to have far ranging macroeconomic implications. First, they provide evidence

that quality and per capita income, by increasing pricing-to-market, play a key role in generating devi-

ations from the Law of One Price, and therefore matter to explain the degree of international market

segmentation. Second, as they improve our understanding of how …rms set their prices when exporting

abroad, and in particular on how they react to exchange rate changes across products di¤erentiated

by quality, they have implications for aggregate pass-through, and therefore for the choice of optimal

monetary and exchange rate policies.7 Third, they carry implications for the creation of currency

unions. As the theory of Optimal Currency Areas emphasizes that the synchronization of in‡ation

shocks between two countries is an important condition for a common currency, our results suggest

that higher income countries trading higher quality goods should be less a¤ected by the in‡ationary

impact of exchange rate depreciations, making a common currency between them more desirable.

Our paper belongs to three strands of the literature. The …rst one is the vast literature on incomplete

exchange rate pass-through and pricing-to-market. These papers usually …nd a low degree of exchange

rate pass-through into import prices (e.g., Campa and Goldberg, 2005; Gopinath and Rigobon, 2008)

or consumer prices (e.g., Campa and Goldberg, 2010). Consistent with other recent …rm-level studies

that use exports data for the whole manufacturing sector (e.g., Amiti et al., 2014; Berman et al., 2012;

Chatterjee et al., 2013), we …nd that exchange rate pass-through for export prices is large.

Within this literature, most closely related to our work are Antoniades and Zaniboni (2013), Auer

and Chaney (2009), and Auer, Chaney, and Sauré (2014) who predict, and show, that pass-through

is higher for lower quality goods (only Auer and Chaney, 2009, do not …nd any evidence for such a

relationship).8 While also focusing on quality and pass-through, we di¤er from these papers along

several dimensions. First, we document the pricing-to-market behavior of exporters that leads to

incomplete pass-through. Instead, Auer et al. (2014) estimate the response of markups to exchange

rate changes at the retail-level, while Antoniades and Zaniboni (2013) and Auer and Chaney (2009)

focus exclusively on exchange rate pass-through. Second, our export prices exclude transportation and

local distribution costs, while the import and retail prices used in these papers contain nontradable

components. Third, while we exploit the within-…rm variation in quality of multi-product …rms, these

papers do not use …rm-level data. Finally, our analysis relies on an observable measure of quality, while

Auer et al. (2014) estimate hedonic quality indices and Antoniades and Zaniboni (2013) and Auer and

Chaney (2009) infer product quality from data on prices and unit values, respectively.

Second, as pricing-to-market and incomplete pass-through lead to deviations from the Law of One

Price, our paper relates to studies that examine the behavior of product-level real exchange rates.

6Compared to the world average, Feenstra and Romalis (2014) estimate that Swiss exports have the highest qualitywhile Chinese exports have the lowest. Across industries, the quality of exports is instead the highest in “Food andbeverages” from France, “Fuels and lubricants” from the US, or “Transport equipment and parts” from Austria.

7As incomplete pass-through determines the extent to which currency changes a¤ect domestic in‡ation in the importingeconomy, it has implications for the implementation of domestic monetary policy. In addition, as incomplete pass-throughdetermines how currency depreciations can stimulate an economy by substituting foreign by domestic goods, it also hasimplications for the evolution of the trade balance and exchange rate policy (Knetter, 1989).

8Basile, de Nardis, and Girardi (2012) develop a model based on Melitz and Ottaviano (2008) and also predict thatpass-through decreases with quality. In contrast, using a translog expenditure function to generate endogenous markupsin a model where …rms are heterogeneous in productivity and product quality, Rodríguez-López (2011) predicts that theresponse of markups to exchange rate shocks decreases with productivity and quality, therefore pass-through increaseswith productivity and quality. In a related study, Yu (2013) shows theoretically that incomplete pass-through resultsfrom …rms adjusting both their markups and the quality of their products in response to a change in the exchange rate.

4

Recent contributions that rely on highly disaggregated product-level data include Simonovska (2014)

who studies the positive relationship between the price of tradable goods and per capita income.

Cavallo, Neiman, and Rigobon (2014) …nd that the Law of One Price holds very well within, but not

outside, of currency unions. Burstein and Jaimovich (2012) show that product-level real exchange rates

across countries are four times as volatile as nominal exchange rates.

Third, this paper contributes to the growing literature on quality and trade, which until recently

mostly relied on trade unit values in order to measure quality.9 At the country-level, Hummels and

Klenow (2005) and Schott (2004) focus on the supply-side, and show that export unit values increase

with exporter per capita income. On the demand-side, Hallak (2006) and Hummels and Skiba (2004)

…nd that richer countries have a stronger demand for high unit value exporting countries. Feenstra and

Romalis (2014) decompose the unit values of internationally traded goods into quality and quality-

adjusted prices, and show that developed countries both export and import more of higher quality

goods. Other papers investigate how quality relates to the performance of exporters using …rm-level

data. Manova and Zhang (2012a) focus on Chinese …rm-level export prices, and …nd evidence of quality

sorting in exports. Kugler and Verhoogen (2012), Manova and Zhang (2012b), and Verhoogen (2008)

highlight the correlation between the quality of inputs and of outputs focusing on Mexican, Chinese,

and Colombian …rms, respectively. Bastos, Silva, and Verhoogen (2014) show that exporting to richer

countries leads Portuguese …rms to upgrade the average quality of their outputs, which requires buying

higher quality inputs. Amiti and Khandelwal (2013) and Fan, Li, and Yeaple (2014) show that lower

import tari¤s are associated with an increase in export quality.10

The paper is organized as follows. Section 2 presents the theoretical model, and discusses how the

features of the wine industry can be reconciled with the main assumptions of the model. Section 3

describes the …rm-level exports customs data, the wine experts quality ratings, and the macroeconomic

data we use. Section 4 presents our main empirical results while Section 5 addresses endogeneity.

Section 6 discusses extensions while Section 7 provides robustness checks. Section 8 concludes.

2 A Model of Pricing-to-Market and Quality

Berman et al. (2012) extend the model with distribution costs of Corsetti and Dedola (2005), allowing

for …rm heterogeneity where single-product …rms di¤er in their productivity. They show that more

productive exporters change their prices more than others in response to a change in the real exchange

rate.11 In their appendix, the authors show that a similar result holds if …rms di¤er in the quality

of the (single) good they export: …rms that export higher quality goods change their export prices

more than others in response to real exchange rate changes. Chatterjee et al. (2013) extend the model

of Berman et al. (2012) to multi-product …rms. They show that in response to a change in the real

exchange rate, exporters vary their prices more for the products they are most e¢cient at producing,

and which therefore have smaller marginal costs.

9This approach is criticized by, among others, Khandelwal (2010) who compares exporters’ market shares conditionalon price to infer the quality of exports.

10Also see Baldwin and Harrigan (2011), Bastos and Silva (2010), Hallak and Schott (2011), Hallak and Sidivasan(2011), and Johnson (2012).

11Berman et al. (2012) obtain similar predictions using the models with endogenous and variable markups of Atkesonand Burstein (2008) and Melitz and Ottaviano (2008).

5

We build on Berman et al. (2012) and Chatterjee et al. (2013) and extend the model of Corsetti

and Dedola (2005), but allow for heterogeneity in the quality of the goods exported. Given that most

…rms in our data set export multiple products, we model them as multi-product …rms which therefore

di¤erentiates us from Berman et al. (2012). In contrast to the multi-product …rms model of Chatterjee

et al. (2013), we rank the di¤erent goods exported by each …rm in terms of quality rather than

e¢ciency, where higher quality is associated with higher marginal costs.12

2.1 The Basic Framework

The Home country (Argentina in our case) exports to multiple destinations in one sector characterized

by monopolistic competition. The representative agent in destination country has preferences over

the consumption of a continuum of di¤erentiated varieties given by13

() =

·Z

ª[()()]

¡1

¸ ¡1

(1)

where () is the consumption of variety , () the quality of variety , and 1 the elasticity

of substitution between varieties.14 The set of available varieties is ª. Quality captures any intrinsic

characteristic or taste preference that makes a variety more appealing for a consumer given its price.

Therefore, consumers love variety but also quality.

Firms are multi-product and heterogeneous in product quality. The parameter , which denotes

each variety, indicates how e¢cient each …rm is at producing each variety, therefore has both a

…rm- and a product-speci…c component. Each …rm produces one “core” product, but in contrast to

Chatterjee et al. (2013) or Mayer, Melitz, and Ottaviano (2014) who consider that a …rm’s core

competency lies in the product it is most e¢cient at producing – and has lower marginal costs – we

assume that a …rm’s core competency is in its product of superior quality which entails higher marginal

costs (Manova and Zhang, 2012b).

The e¢ciency associated with the core product is given by a random draw ©, therefore each …rm

is indexed by ©. Let us denote by the rank of the products in increasing order of distance from the

…rm’s core, with = 0 referring to the core product with the highest quality. Firms therefore observe

a hierarchy of products based on their quality levels. A …rm with core e¢ciency © produces a product

with an e¢ciency level given by

(© ) = © (2)

where 1. Products with smaller (higher quality) are closer to the core, and have a lower e¢ciency

12Higher quality goods are typically assumed to have higher marginal costs because they require higher quality – andtherefore more expensive – inputs. See Crinò and Epifani (2012), Fan, Li, and Yeaple (2014), Feenstra and Romalis(2014), Hallak and Sivadasan (2011), Johnson (2012), Kugler and Verhoogen (2012), Manova and Zhang (2012a), andVerhoogen (2008).

13For similar preferences, see Baldwin and Harrigan (2011), Berman et al. (2012), Crozet et al. (2012), Fan et al.(2014), Johnson (2012), Kugler and Verhoogen (2012), and Manova and Zhang (2012b).

14As in Kugler and Verhoogen (2012), quality is treated as single-dimensional and is therefore only chosen by …rms.

6

(© ). Higher quality goods have a lower e¢ciency because they have higher marginal costs15

( (© )) =

µ

(© )

¶

(3)

where 1 implies that markups increase with quality, and is the wage of the Home country

(Berman et al., 2012).16 The closer a product is from the core with the highest quality (i.e., the

smaller ), the lower is e¢ciency (© ), and the higher are marginal costs and quality ( (© )).

Firms face three types of transaction costs: an iceberg trade cost 1 (between Home and

destination ), a …xed cost of exporting (which is the same for all …rms and products and only

depends on destination ), and an additive (per unit) distribution cost in destination .17 The latter

captures any wholesale and retail costs associated with the selling, storage, marketing, advertising, or

transport of the goods that are paid in the currency of the destination country.18 If distribution requires

units of labor in country per unit sold, and is the wage rate in country , distribution costs are

given by ((© )). As in Berman et al. (2012), we assume that higher quality goods have higher

(per unit) distribution costs. Most importantly, as distribution is outsourced so that distribution costs

are paid in the currency of the importing country, they are una¤ected by changes in the exchange rate.

In units of currency of country , the consumer price in of a variety exported from Home to is

() ´((© ))

+ ((© )) (4)

where () is the export price of the good exported to , expressed in Home currency, and is

the nominal exchange rate between Home and . It is straightforward to see that any change in the

exchange rate leads to a less than proportional change in the consumer price () (i.e., incomplete

pass-through) given that local distribution costs are una¤ected by currency ‡uctuations. The quantity

demanded for this variety in country is

() = ¡1

·((© ) ((© ))

+

¸¡ (5)

where and are country ’s income and aggregate price index, respectively.19 The costs, in the

currency of the Home country, of producing () units of each good (inclusive of transportation

costs) and of selling them to country are

() = ((© ))

(© )+ (6)

15See Crinò and Epifani (2012) and Hallak and Sivadasan (2011) for models where marginal costs decrease with …rmlevel productivity and increase with product quality.

16Auer et al. (2014) observe that producers charge a higher markup when exporting higher quality cars.17We do not model entry and exit in export markets. See Campos (2010) for the e¤ects of entry and exit on pass-through.18Local distribution costs are economically important. Burstein et al. (2003) show that distribution costs represent

between 40 and 60 percent of retail prices across countries. Campa and Goldberg (2010) show that local distributioncosts, which represent between 30 and 50 percent of the total costs of goods exported by 21 OECD countries in 29manufacturing industries, decrease exchange rate pass-through into import prices. Hellerstein (2008) shows that partialpass-through can be explained by markup adjustments and by the presence of local costs in roughly similar proportions.

19The aggregate price index is given by =ª ()

1¡ 1

1¡ .

7

Expressed in Home currency, the pro…t maximizing export price for each product the …rm exports to

country is

() =

¡ 1

µ

1 +(© ) ((© ))

¶

(© )= ((© ))

(© ) (7)

where ´ is the real exchange rate between Home and . In contrast to the standard Dixit-

Stiglitz markup (Dixit and Stiglitz, 1977), the presence of local distribution costs leads to variable

markups ((© )) over marginal costs that are larger than ¡1 , increase with quality ((© )),

the real exchange rate (i.e., a real depreciation), and local distribution costs .

The volume of exports () is given by

() =

µ ¡ 1

¶

¡1

·

(© )((© )) +

¸¡(8)

therefore the elasticity (in absolute value) of the exporter’s demand () with respect to the export

price () is

=

¯¯¯¯()

()

()

()

¯¯¯¯ =

+ (© )((© ))

+ (© )((© )) (9)

which is decreasing in quality and with a real depreciation. For a product that is closer to the core,

quality is higher, the elasticity of demand is smaller, and the markup is higher. The model leads to

two testable predictions on the e¤ects of exchange rate changes on export prices and quantities.

Prediction 1 The …rm- and product-speci…c elasticity of the export price () to a change in the

real exchange rate , denoted by and which captures the degree of pricing-to-market, increases with

the quality of the good exported, ((© )):

=

¯¯¯¯()

()

¯¯¯¯ =

(© )((© ))

+ (© )((© ))

Prediction 2 The …rm- and product-speci…c elasticity of the volume of exports () to a change in

the real exchange rate , denoted by , decreases with the quality of the good exported, ((© )):

=

¯¯¯¯()

()

¯¯¯¯ =

+ (© )((© ))

Intuitively, the mechanism is the following. A real depreciation reduces the elasticity of demand

perceived by exporters in the destination country, which allows all …rms to increase their markups. As

higher quality goods have a smaller elasticity of demand, their markups can therefore be increased by

more than for lower quality goods. This leads to heterogeneous pricing-to-market which is stronger

for higher quality goods (i.e., pass-through is lower). In turn, this implies that the response of export

volumes to a real depreciation decreases with quality. This mechanism is similar to Berman et al.

(2012) and Chatterjee et al. (2013), although their focus is on productivity di¤erences in driving

heterogeneous pricing-to-market across exporters, or exporters and products, respectively.

8

2.2 Cross-Country Heterogeneity in the Preference for Quality

The evidence in the literature suggests that consumer preferences for quality vary from one country to

the other as preferences are a¤ected by per capita income. In particular, consumers in richer countries

are expected to have a stronger preference for higher quality goods.20 We therefore extend the model of

the previous section and allow for nonhomothetic preferences for quality. The utility function becomes

() =

·Z

ª[( )()]

¡1

¸ ¡1

(10)

where we replace () by ( ) to capture that the preference for quality can increase in per capita

income, . Local distribution costs ( ) are thus higher in higher income countries.21 This

allows us to derive two additional testable predictions.

Prediction 3 The …rm- and product-speci…c elasticity of the export price () to a change in the

real exchange rate , denoted by , increases with the quality of the good exported ((© ) ), and

by more for higher income than for lower income destination countries:

=

¯¯¯¯()

()

¯¯¯¯ =

(© ) ((© ) )

+ (© ) ((© ) )

Prediction 4 The …rm- and product-speci…c elasticity of the volume of exports () to a change in

the real exchange rate , denoted by , decreases with the quality of the good exported ((© ) ),

and by more for higher income than for lower income destination countries:

=

¯¯¯¯()

()

¯¯¯¯ =

+ (© ) ((© ) )

The elasticity of the exporter’s demand with respect to the export price is now decreasing in quality,

per capita income, and with a real depreciation. In response to a real depreciation, …rms can therefore

increase their markups by more for higher quality goods, but also when exporting to higher income

destination countries.22 In turn, this implies that the response of volumes to exchange rate changes

decreases with quality and per capita income. The assumption that consumers in richer countries are

less price-elastic than consumers in poorer countries – which enables …rms to charge higher markups

in richer destinations – is consistent with Alessandria and Kaboski (2011) and Simonovska (2014).

2.3 Relevance for Wine

The predictions of our model, which features monopolistic competition and local distribution costs,

depend crucially on the assumptions that higher quality wines have higher (per unit) distribution costs,

20Crinò and Epifani (2012) …nd that the preference for quality is on average 20 times larger in the richest (the US) thanin the poorest location (Africa) in their sample. Hallak (2006) …nds that rich countries tend to import relatively morefrom countries that produce higher quality goods. Fajgelbaum, Grossman, and Helpman (2011) develop a model whereconsumers have heterogeneous incomes and tastes, and the share of consumers who buy higher quality goods increases inincome. Verhoogen (2008) assumes that Northern consumers are more willing to pay for quality than Southern consumers.Also see Feenstra and Romalis (2014), Jaimovich and Merella (2014), Manova and Zhang (2012a), and Simonovska (2014).

21Dornbusch (1989) shows that the prices of services are lower in poorer than in richer countries, suggesting that localdistribution costs are lower in poorer countries.

22 In Auer et al. (2014), pass-through decreases with quality in rich countries and rises with quality in poor countries.

9

higher marginal costs, and higher markups. This section discusses how the characteristics of the wine

industry can be reconciled with these assumptions.

In terms of market structure, the wine industry is composed of many producers o¤ering a large

number of highly di¤erentiated products for which consumers are willing to pay a wide range of di¤erent

prices. Because of product di¤erentiation, wine producers have market power. Therefore, the sector can

legitimately be assumed to operate in a monopolistically competitive environment (Thornton, 2013).

Our model features the role of local distribution costs in generating incomplete pass-through. In the

case of wine, local distribution costs are indeed important. For instance, in the US, about 90 percent of

wine is distributed through a three-tier distribution system, whereby wine is sold by a producer or by

an importer to a distributor, who then sells it to a retailer, who …nally sells it to consumers. Each tier

involves a markup to compensate for each activity that is carried out. The markup of the distributor is

usually around 30 percent, and covers the costs related to transport, storage, and promotional activities.

In turn, the markup of the retailer varies between 30 percent for grocery stores and wine shops and

200 percent for restaurants, bars, and hotels (Thornton, 2013).

The prediction of our model that pass-through decreases with quality depends crucially on the

assumption that higher quality goods have higher (per unit) distribution costs. In the case of wine,

this assumption is plausible. First, higher quality wines are more costly to transport as they are typi-

cally shipped in temperature-controlled containers, they need heavier bottles and cartons for shipping,

require more careful and therefore more costly handling, and face higher insurance fees (Hummels and

Skiba, 2004). Second, tasting events to promote wines are organized by distributors for higher quality

wines only, which therefore have higher marketing and advertising costs. Third, top quality wines need

to be stored in refrigeration units to maintain their temperature, and are therefore subject to higher

storage costs. Finally, top quality wines are more likely to be sold in specialized wine stores than in

supermarkets, and the former usually charge a higher price because of the professional customer service

they o¤er. In other words, retail costs are higher for top quality wines.

Higher quality wines can also be expected to entail higher marginal costs (Thornton, 2013). First,

the quality of wine depends predominantly on the quality of the grapes. Although grape quality is

mostly a¤ected by geography and weather-related factors, achieving higher quality grapes can be more

costly as it requires, for instance, to restrict the grape yield per acre of vineyard, to grow the grapes

organically, to use labor-intensive treillis systems, and to prune, trim, pick, and sort the grapes more

carefully which in turn necessitates more skilled, manual labor. Higher quality wines also require

higher quality, and therefore more costly inputs such as the additives and yeast types used during the

various stages of fermentation or as preservatives. Second, achieving higher quality wines depends on

the production methods chosen, including the way in which the wine is extracted from the grapes (the

manual labor-intensive “punching down” technique is more costly but results in superior wine quality),

or the type of vessel to be used for ageing and fermentation. Compared to stainless-steel tanks, oak

barrels increase quality by adding ‡avor to the wine, but are much more expensive due to the cost of

the oak (the barrels are between 75 and 90 percent more costly than similar sized stainless-steel tanks,

and add between two to three US dollars to the cost of a bottle of wine), their short lifetime (the oak

‡avors last for three or four vintages only), and the evaporation of the wine due to the porous nature

of the oak (between 10 to 12 percent per year). Quality can be further enhanced by using smaller oak

10

barrels to increase the amount of wood per unit of wine, but these are in turn more expensive. Finally,

higher quality wines need to mature for a longer period of time, which increases storage costs.

As an illustration, Table 1 breaks down into several components the prices of non-EU wines sold

in UK retail outlets (Joseph, 2012). We believe that these …gures should provide us with some useful

insights on the composition of Argentinean wine prices sold in the UK. The table shows that more

expensive, and hence higher quality wines have higher retailing and distribution costs. Also, the

amount that goes to the winemaker, which includes his pro…t as well as the costs of producing the

wine, clearly increases with the price, and therefore with quality. Unfortunately, the table does not

provide any information on the winemaker markup which is the one that is modeled in the theory.

However, anecdotal evidence suggests that the producer markup is also likely to increase with the

price/quality of the wine: for a …ve pound wine sold on the UK market, the producer markup is

estimated to be around 40 pence, and increases to about ten pounds for a 25 pound bottle.23

3 Data and Descriptive Statistics

Our data set gathers information from di¤erent sources: …rm-level customs data, wine experts quality

ratings, and macroeconomic data.

3.1 Firm-Level Customs Data

Our …rm-level exports data are from the Argentinean customs and are provided to us by a private

vendor called Nosis. For each export ‡ow we have the name of the exporting …rm, the country of

destination, the date of declaration, the 12-digit HS classi…cation code, the FOB value of exports (in

US dollars), and the volume (in liters) exported between 2002 and 2009.24 We also have the name of

the wine exported, its type (red, white, or rosé), grape (Malbec, Chardonnay, etc.), and vintage year.

Figure 1 compares the total value of Argentina’s wine exports from the customs data set with the value

reported in the Commodity Trade Statistics Database (Comtrade) of the United Nations (HS code

2204). The data coincide extremely well.

Given that actual export prices are not available, we proxy for them using the unit values of exports

in the currency of the exporter, computed as the ratio of the export value in Argentinean pesos divided

by the corresponding export volume in liters. In order to convert the value of exports (in US dollars)

into pesos, we use the peso to US dollar exchange rate in the month in which the shipment took place.

We then aggregate the data at an annual frequency.25

23See www.thirty…fty.co.uk.24Due to con…dentiality reasons imposed by Argentinean law, the customs data cannot make the name of the exporter

public. However, Nosis uses its own market knowledge to generate a “…rst probable exporter,” a “second probableexporter,” and a “third probable exporter.” To determine the exporter’s identity, we used from the Instituto Nacionalde Vinticultura (INV) the names of all wines and of all the …rms authorized to produce and sell wine and compared,for each wine name, the name of the …rst probable exporter with the authorized exporter reported by the INV. If thisname coincided, we kept the …rst probable exporter. Otherwise, we repeated the procedure with the second probableexporter, and …nally with the third probable exporter. We identi…ed 86 percent of the …rms as “…rst probable exporters,”…ve percent as “second probable exporters,” one percent as “third probable exporters,” while the remaining eight percentcould not be matched. In our sample, these proportions become 96.5, three, one half, and zero percent, respectively.

25Our results still hold when measuring unit values, export volumes, and exchange rates at a quarterly frequency. Thereason why we aggregate the data at an annual frequency is that many of the other variables we use – such as the reale¤ective exchange rates or real GDPs per capita – are not available at a quarterly frequency.

11

We clean up the data in several ways. First, we drop any wine for which either the name, grape,

type, or vintage year is missing, cannot be recognized, or is classi…ed as “Unde…ned.” As sparkling

wines, dessert wines, and other special varieties do not have any vintage year, they are excluded from

the data set. Second, we only keep the export ‡ows recorded as FOB, which therefore exclude shipping

costs, tari¤s, or distribution and retail costs charged in the destination country. Third, as we are

interested in how product quality a¤ects the pricing and export decisions of wine producers, we restrict

our analysis to the manufacturing sector, and drop wholesalers and retailers. Fourth, due to the rise in

in‡ation in 2007 and increasing worries of capital out‡ows, the government introduced some barriers

to obtain import licenses which culminated in an extreme measure whereby …rms were authorized to

import if they exported goods for an equivalent value to those they intended to import.26 We therefore

drop from the sample the …rms unrelated to the wine industry that exported wine over the period.27

Fifth, we drop the wines that are not produced in Argentina.28 Sixth, we drop a number of typos which

we are unable to …x. We exclude the few cases where the vintage year reported is ahead of the year in

which the export took place. We also drop the observations where the value of exports is positive, but

the corresponding volume is zero. Finally, we also exclude some outliers: for each exporter, we drop

the observations where unit values are larger or smaller than 100 times the median export unit value

charged by the …rm.

The recent papers on heterogeneous pass-through typically de…ne a “product” according to trade

classi…cations such as the CN or the HS (e.g., Amiti et al., 2014; Auer and Chaney, 2009; Berman et

al., 2012; Chatterjee et al., 2013). As Table 2 shows, the 6-digit HS classi…es wines into four di¤erent

categories, depending on whether the wines are sparkling or not, and on the capacity of the containers

in which they are shipped (i.e., larger or smaller than two liters). Argentina further disaggregates

the HS classi…cation at the 12-digit level, but this only enlarges the number of di¤erent categories,

or “products,” to eleven. The problem is that changes in unit values de…ned at this level may re‡ect

compositional or quality changes rather than price changes as there may be more than one distinct

product within a single HS code (Feenstra and Romalis, 2014). In contrast, the detail provided by our

data set allows us to de…ne an individual product as a combination between a wine name, type, grape,

vintage year, and the capacity of the container used for shipping (identi…ed using the HS code), so that

compositional or quality changes are unlikely to a¤ect unit values.29 Our cleaned sample includes a

total of 21,647 di¤erent products/wines of which 6,720 can be matched with quality ratings. The 6,720

wines represent 43 percent of the total FOB value of red, white, and rosé wine exported between 2002

and 2009.

26See http://www.ustr.gov/sites/default/…les/2013%20NTE%20Argentina%20Final.pdf. The World Trade Organiza-tion has subsequently ruled that the requirement that imports should be balanced with exports violates global trade rules(The Financial Times, “WTO Rules Against Argentine Import Curbs,” 22 August 2014).

27We discard a total of 157 …rms that include wine wholesalers (20), wine retailers (2), unidenti…ed …rms (70), as wellas …rms that belong to the sectors of agriculture (5), soft drinks or malt liquors (6), textiles (6), chemicals (6), metals(1), paints, glass, and paper (3), machinery (1), motorcycles (1), and services industries (25) such as consultancy or bookpublishing …rms. These …rms represent 12,376 observations in the raw data set (at the transaction-level), or 1.6 percentof the total FOB value of wine exported by Argentina between 2002 and 2009. Of the 157 …rms, only 41 export winesthat can be matched with the quality ratings from the Wine Spectator, and these …rms represent 0.17 percent of the totalFOB value exported over the period (or 1,520 observations). In the …nal data set aggregated at a yearly frequency, the41 …rms represent 345 observations.

28These wines represent 0.7 percent of the total FOB value of wine exported between 2002 and 2009.29 In our sample, each wine is produced, and exported, by one …rm only. In contrast, a product de…ned at the CN or

HS levels can be produced, and therefore exported, by more than one …rm.

12

3.2 Quality Ratings

The editors of the Wine Spectator magazine review more than 15,000 wines each year in blind tastings,

and publish their ratings in several issues throughout the year. The ratings are given on a (50,100)

scale according to the name of the wine, its grape, type, and vintage year, which are characteristics

we all observe in the customs data set. A larger score implies a higher quality.30. Table 3 lists the six

di¤erent categories the wines fall in depending on their scores.

We match the wines from the customs data set with the ones reviewed by the Wine Spectator

by name, type, grape, and vintage year, and each wine is assigned a single quality rating.31 We end

up with 6,720 wines exported by 209 …rms over the 2002-2009 period. Our approach to measuring

quality is similar to Crozet et al. (2012) for Champagne. However, they are unable to distinguish

between the di¤erent varieties sold by each …rm, and therefore assume that each …rm exports one type

of Champagne only. In addition, their ratings are only measured on a (1,5) scale, with a larger value

indicating a higher quality.

We rely on the Wine Spectator for our baseline regressions because it has the largest coverage of

Argentinean wines. However, in the robustness section, we check the sensitivity of our results using an

alternative rating produced by Robert Parker. Parker is a leading US wine critic who assesses wines

based on blind tastings, and publishes his consumer advice and ratings in a bimonthly publication,

the Wine Advocate.32 His rating system also employs a (50,100) point scale, with wines being ranked

according to their name, type, grape, and vintage year, and with a larger value indicating a higher

quality.33 Table 3 lists the di¤erent categories considered by Parker. Compared to the Wine Spectator,

the scores are slightly more generous (for instance, a wine ranked 74 is “Not recommended” by the

Wine Spectator, but is “Average” according to Parker). The customs data and the Parker ratings can

be matched for 3,969 wines exported by 181 …rms.

Figure 2 plots the Wine Spectator against the Parker ratings for the 2,433 wines (exported by 135

…rms) that have scores from both sources. For the sake of clarity, for both ratings quality is measured

between one and six, where each value corresponds to one of the di¤erent bins de…ned in Table 3. A

value of one indicates that the wine is either “Not recommended” or “Unacceptable,” while a value of

six that the wine is “Great” or “Extraordinary.” For Parker, notice that only the three highest quality

bins are depicted in the graph.34 Each circle is sized proportionately to the number of wines at that

30Quality depends on the amounts of characteristics embodied in a wine, and is assessed by evaluating those char-acteristics. Subjective quality refers to the evaluation of consumers based on personal tastes and preferences, whereasobjective quality assesses the characteristics that are independent of personal preferences such as balance, complexity,…nish, concentration, or ‡avor intensity. The aim of wine critics is to provide an objective assessment of wine quality,and they may therefore conclude that a wine is of high quality even though they personally dislike it (Thornton, 2013).However, as we discuss later, some argue that experts ratings can, to some extent, be subjective, and therefore re‡ect thepersonal preferences of the tasters.

31The quality scores are time-invariant. Although some evidence suggests that the quality of Argentinean wines (andtheir price) is not much a¤ected by ageing (Thornton, 2013), in some of our regressions we include vintage year …xede¤ects to control for product characteristics. When explaining unit values, the coe¢cients on the vintage year …xed e¤ectsare larger for older wines, indicating that older wines tend to be more expensive. A higher price for older wines couldcapture that quality improves with ageing, but may also re‡ect the costs incurred by producers for storing the wines ininventories for several years.

32Both the Wine Spectator and Parker are US-based ratings. Other ratings have a poor coverage of Argentinean wines.33Starting with a base score of 50, Parker assigns additional points for appearance (5 points), smell (15 points), taste

and …nish (20 points), and overall quality and ageing potential (10 points). Instead, the Wine Spectator does not makepublic how the points are awarded.

34When using Parker in our regressions, four out of the six bins listed in Table 3 are however included in our sample.

13

point in the …gure. Most wines are rated as “Above average/very good” (a value of four) by Parker,

and are either “Good” or “Very good” (a value of three or four) according to the Wine Spectator,

again suggesting that Parker is, on average, more generous in its ratings. A relatively large number

of wines are rated as “Outstanding” (a value of …ve) by both tasters, and some as either “Great” or

“Extraordinary” (a value of six). Notice that a few wines are ranked as “Not recommended” (a value

of one) by the Wine Spectator, but “Outstanding” (a value of 5) according to Parker. The two ratings

are, however, positively correlated. The Pearson correlation is equal to 0.51, while Kendall’s correlation

index of concordance between the two ratings is 0.47.

3.3 Macroeconomic Data

The data on real GDPs per capita (in US dollars) and labor compensation as a share of GDP are from

the Penn World Tables, and the consumer price indices (CPI) and nominal exchange rates from the

International Financial Statistics (IFS) of the International Monetary Fund (IMF). The real exchange

rate is de…ned as the ratio of consumer price indices times the average yearly nominal exchange rate,

and an increase indicates a real depreciation of the Argentinean peso. The real e¤ective exchange

rates are sourced from the IFS and the Bank of International Settlements, and an increase indicates

a real depreciation. Bilateral distances are from the Centre d’Etudes Prospectives et d’Informations

Internationales (CEPII). Real wages are obtained by multiplying the labor compensation share by the

destination country’s GDP, and de‡ating using the corresponding CPI.

During the 2002-2009 period, Argentina witnessed major nominal and real exchange rate ‡uctua-

tions. Figure 3 illustrates the evolution of the monthly nominal and real exchange rates between the

Argentinean peso and the US dollar. After the …nancial crisis of 2001, the …xed exchange rate system

was abandoned and as a result, in 2002, the peso depreciated both in nominal and real terms by up to

75 percent. The export boom that followed lead to a massive in‡ow of US dollars into the economy,

which helped to appreciate the peso against the dollar. The peso then remained stable – but slightly

appreciated in real terms – until 2008 when it depreciated again with the advent of the global …nancial

crisis and the increase in domestic in‡ation.

3.4 Descriptive Statistics

Table 4 summarizes our trade data by year, and shows that the exports included in our sample increased

threefold between 2002 and 2009. A total of 794 wines were exported by 59 di¤erent …rms in 2002,

while in 2009 this increased to 151 …rms exporting 1,833 di¤erent wines. Over the whole period, our

sample includes 6,720 wines exported by 209 di¤erent wine producers.35 As shown in Table 5, these

…rms exported an average of 139 di¤erent wines, from a minimum of one to a maximum of 510 (in the

sample, 15 …rms appear as having exported one wine only; in reality, they exported more than one

wine, but only one could be matched with the quality ratings). Exporters charged between two cents

and 381 US dollars per liter of wine exported, with an average of …ve US dollars per liter.36 They

exported to an average of 24 di¤erent destinations, from a minimum of one to a maximum of 65. The

35We observe 882 di¤erent wine names, 23 grapes, three types, and 22 vintage years (between 1977 and 2009).36The Trade Unit Values Database developed by the CEPII uses the bilateral trade values and quantities collected by

the United Nations to compute “reliable and comparable unit values across countries” (Berthou and Emlinger, 2011, p.3).At the 6-digit level of the HS classi…cation, the FOB unit values of Argentinean wine exports between 2002 and 2009 varybetween 7 cents and 16 US dollars per liter exported. Given the much …ner disaggregation level of our data, unit valuesbetween 2 cents and 381 US dollars per liter sound reasonable.

14

mean Wine Spectator rating is 85, the lowest rated wine receives a score of 55, and the highest receives

a score of 97. The distribution across wines is symmetric as the mean and the median are equal. The

Parker scores vary between 72 and 98, with an average of 87 (only four out of the six bins listed in Table

3 are therefore included in our sample). Table 6 shows that, with the exception of Brazil, Argentinean

wine producers mostly exported to developed economies, the United States being the top destination

market.

Table 7 provides a snapshot of our data. For con…dentiality reasons, we can neither report the

exporter nor the wine names, therefore these are replaced by numbers and letters instead. The table

shows that, whether we use the Wine Spectator or the Parker ratings, individual …rms exported wines

with heterogeneous levels of quality (between 73 and 92 for Firm 1, and between 85 and 98 for Firm 2).

In addition, higher quality wines are, on average, sold at a higher price. Besides, the table illustrates

that the Law of One Price fails: in 2006, Firm 1 exported the same wine to two di¤erent destinations,

but charged 13.38 US dollars per liter to the United States, versus 11.09 US dollars per liter to China.

Similarly, in 2008, Firm 2 charged 43.27 US dollars to the United Kingdom, versus 24.65 US dollars

to Thailand for the same liter of wine exported. The di¤erences in the prices of these wines re‡ect

di¤erences in markups as identical wines have the same marginal cost. Interestingly, the prices, and

therefore the markups, are higher for exports to higher income countries, which is consistent with

Simonovska (2014) who argues that exporters extract higher markups for identical products when

selling to richer countries with lower demand elasticities.

Table 8 reports descriptive statistics by quality bin of the Wine Spectator. Wines that are either

“Good” or “Very good” represent the largest share of the sample (in terms of the total number of

observations, the number of …rms exporting these wines, and the share of exports in the sample).

At the other two extremes, “Great” wines, followed by “Not recommended” wines, have the smallest

coverages. Compared to other wines, “Great” wines are more expensive, and are exported to a smaller

number of destinations (per wine exported) which are, on average, richer (Hallak, 2006; Verhoogen,

2008), and more distant (Baldwin and Harrigan, 2011; Bastos and Silva, 2010; Feenstra and Romalis,

2014; Hummels and Skiba, 2004; Johnson, 2012).

Finally, Table 9 describes the data across …rms. Each …rm is allocated to one of the Wine Spectator

quality bins, based on the highest quality wine the …rm exported over the period. Firms exporting

higher quality wines tend to export a higher quality, on average.37 Interestingly, these …rms also tend

to export more variety. Taken together, the 68 …rms which exported “Great” and “Outstanding” wines

exported “Not recommended” wines as well. They exported a larger number of di¤erent wines, which

are shipped to a larger number of destinations (per wine exported), and represent the largest share of

exports in the sample. At the other extreme, only one …rm exported exclusively “Not recommended”

wines, while 12 …rms exported “Mediocre” wines as their highest quality.

4 Empirical Analysis

In order to check if Prediction 1 holds in the data, we estimate the following reduced-form regression

ln = 1 ln + 2 ln £ + + + (11)

37A …rm can share its expertise in winemaking across all of its products, raising their average quality (Thornton, 2013).

15

where is the export unit value of wine exported by …rm to destination country in year ,

expressed in pesos per liter of wine exported, and is our proxy for export prices. is the average

real exchange rate between Argentina and country in year (an increase in indicates a real

depreciation). The quality of wine is denoted by , and the Wine Spectator ratings are used

for our benchmark speci…cations.

The export price in the exporter’s currency is a markup over marginal costs. In order to identify a

pricing-to-market behavior, which requires markups to respond to exchange rate changes, the regression

needs to control for product-speci…c marginal costs (Knetter, 1989, 1993).38 As we assume in the model

of Section 2, higher quality wines can be expected to have higher marginal costs. We therefore perform

within estimations, and control for time-varying product-speci…c marginal costs (which are, as a matter

of fact, also …rm-speci…c) by including product-year …xed e¤ects, .39 As product-year …xed e¤ects

are collinear with quality, the direct e¤ect of quality drops out from the regression. Firm-destination

…xed e¤ects, , are also included. The coe¢cients to be estimated are 1 and 2 and is an error

term. Prediction 1 requires 2, the coe¢cient on the interaction between the real exchange rate and

quality, to be positive.40

As all variables are in levels (rather than …rst di¤erences), the estimated coe¢cients can be thought

of as capturing the long term response of unit values to changes in each of the explanatory variables.

In addition, given the level of disaggregation of the data, changes in real exchange rates are assumed to

be exogenous to the pricing decisions of individual …rms.41 Finally, as quality takes on a single value

for each product, robust standard errors are adjusted for clustering at the product-level.

To test Prediction 2 we estimate

ln = 1 ln + 2 ln £ + 3 + + + (12)

where is the FOB export volume (in liters) of wine exported by …rm to destination country

in year . To be consistent with gravity models, we include destination-year-speci…c variables ,

including the importing country’s real GDP per capita and real e¤ective exchange rate as a proxy for

country ’s price index (Berman et al., 2012). Prediction 2 requires the coe¢cient on the interaction

term, 2, to be negative.

4.1 Baseline Results

Panel A of Table 10 reports the results for export unit values. In order to get a sense of how quality

a¤ects unit values, and therefore allow the coe¢cient on quality to be estimated in the regression, we

…rst estimate equation (11) by replacing the product-year …xed e¤ects by …rm-year dummy variables

38Distinguishing between changes in markups and changes in marginal costs is also important from a policy point ofview as changes in costs are e¢cient, while changes in markups are not (Burstein and Jaimovich, 2012).

39As each wine is only produced in a single vintage year, product-speci…c marginal costs are time-invariant. Still, weinclude product-year …xed e¤ects to control for any other time-varying costs that may di¤er across products, and thereforeacross di¤erent qualities, such as storage costs.

40When the interaction between the exchange rate and quality is excluded from (11), 1 captures the degree of pricing-to-market. Exchange rate pass-through for export prices is given by (1¡ 1)£ 100.

41Berman et al. (2012) estimate a similar regression in levels. Other papers focus on …rst di¤erences in order to addressnon-stationarity (e.g., Gopinath et al., 2010; Gopinath and Rigobon, 2008). Given the level of disaggregation of our data,computing …rst di¤erences in unit values reduces our sample size by four, and our estimates become insigni…cant.

16

(to control, among other factors, for the time-varying marginal costs or reputation of each exporter),

and simultaneously control for product characteristics by including grape, type, vintage year, HS, and

province of origin of the grapes …xed e¤ects.42 Fixed e¤ects for the wine names are not included as

they are collinear with the …rm …xed e¤ects (as each wine is exported by one producer only). Column

(1) only includes the exchange rate and quality as regressors, and shows that higher quality wines are

exported at a higher price, which is consistent with equation (7) and with the …ndings of Crozet et al.

(2012) for Champagne. Also, as expected, exporters increase their export prices in response to a real

depreciation.

Column (2) includes product-year …xed e¤ects to control for the time-varying marginal costs of

producing each wine, therefore quality drops out from the regression. When the real exchange rate

‡uctuates, exporters change their export prices: in response to a ten percent real depreciation, they

raise their prices (in pesos) by 1.9 percent, therefore pricing-to-market is low and pass-through for

export prices is high at 81 percent. This contrasts with the …ndings of papers that use import rather

than export prices, e.g., Gopinath and Rigobon (2008) …nd that the pass-through for import prices in

the US is only 22 percent.43 The large degree of pass-through we …nd for the wine industry is, however,

consistent with the …ndings of other papers that use …rm-level exports data for the whole manufacturing

sector. For instance, pass-through for export prices is estimated at 79 percent for Belgian exporters

(Amiti et al., 2014), 92 percent for French exporters (Berman et al., 2012), 77 percent for Brazilian

exporters (Chatterjee et al., 2013), 86 percent for Danish exporters (Fosse, 2012), and at 94 percent

for Chinese exporters (Li, Ma, Xu, and Xiong, 2012).44

The estimated coe¢cient on the exchange rate in column (2) however hides a signi…cant amount

of heterogeneity in the degree of pass-through across products. To see this, column (3) interacts the

exchange rate with quality. Its estimated coe¢cient is positive and signi…cant, providing support to

Prediction 1 that the elasticity of export prices to a change in the real exchange rate increases with

quality – in other words, pass-through decreases with quality. The latter …nding is consistent with the

theoretical predictions of Auer and Chaney (2009) and Basile, de Nardis, and Girardi (2012), and with

the empirical results of Antoniades and Zaniboni (2013) and Auer et al. (2014).45

Both quality and pass-through may vary independently from each other over time. We therefore

need to ensure that the negative relationship between quality and pass-through that we …nd is not

42“Old World” wines produced in Europe are labelled by the region where they are made and where the grapes aregrown. In contrast, wines in Argentina follow the tradition of “New World” wines and are classi…ed by the grape varietyused for producing the wine, and the grapes can be grown in di¤erent provinces. This is the reason why, in our regressions,we control for the province of origin of the grapes. According to Thornton (2013), the location where the grapes are grownmay also a¤ect wine prices through its reputation.

43One possible reason for low pricing-to-market, and therefore for high pass-through, is transfer pricing between multi-national …rms exporting to foreign subsidiaries. If markups are not adjusted by exporters but by their subsidiaries inthe destination country, pass-through can be high for export prices but low for consumer prices (Knetter, 1992). Thisscenario is unlikely in our case as our sample includes wine producers only and no multinational …rms.

44We checked if exporters adopt di¤erent pricing strategies depending on whether the Argentinean peso appreciates ordepreciates in real terms. We found that unit values respond signi…cantly more to appreciations than to depreciations.This asymmetric pattern is consistent with …rms trying to maintain export market shares by reducing the domesticcurrency prices of their exports which become less competitive when the peso appreciates, and with Marston (1990) whoshows that pricing-to-market by Japanese …rms is stronger when the Japanese yen appreciates.

45 In response to a change in the exchange rate, multi-product …rms could also modify the composition of their exportsand therefore change the number of products they sell abroad (see Berman et al., 2012, for a discussion, and Chatterjeeet al., 2013, for some evidence). One way to address this possibility is to restrict the sample to …rms exporting only onewine to each destination. This reduces our sample to 341 observations, and our estimates become insigni…cant.

17

driven by pass-through being low when quality is high for unrelated reasons. To address this possibility,

column (4) interacts the real exchange rate with year dummies. Reassuringly, the coe¢cient on the

interaction between the exchange rate and quality remains positive and signi…cant.

Column (5) replaces the …rm-destination …xed e¤ects by …rm-destination-year dummies to control

for excluded characteristics that vary by …rm-destination-year (such as the time-varying demand of a

country for a …rm’s exports, or the presence of long term contracts between exporters and importers in

each destination country). In column (6), we estimate a di¤erence-in-di¤erence speci…cation by includ-

ing destination-year, …rm-destination, and product-year …xed e¤ects. In both cases the exchange rate

drops out, but the interactions between the exchange rate and quality remain positive and signi…cant.46

As shown in Figure 3, Argentina has experienced two major exchange rate regime shifts during the

sample period. First, after the …xed exchange rate between the Argentinean peso and the US dollar

was abandoned in 2001, the peso depreciated greatly with respect to the US dollar throughout 2002.

Second, with the …nancial crisis that started in 2008, the peso depreciated again with respect to the

US dollar. In addition, the crisis might have prompted consumers to substitute towards lower quality

goods (a “‡ight from quality,” see Burstein, Eichenbaum, and Rebelo, 2005; Chen and Juvenal, 2015),

and impacted the …nancial constraints of exporters (Strasser, 2013). Columns (7) and (8) therefore

restrict the sample to the post-2002 and pre-2008 periods, respectively, while column (9) focuses on

the 2003-2007 period only. Our results continue to hold.

Panel B reports the results of estimating equation (12) for export volumes. In column (1), where

the product-year …xed e¤ects are replaced by …rm-year, grape, type, vintage year, HS, and province

of origin dummy variables, the coe¢cient on quality is negative and signi…cant, which at …rst glance

contrasts with the positive relationship between trade and quality that is usually estimated in the

literature (e.g., Crozet et al., 2012). One crucial di¤erence between our regression and, for instance,

Crozet et al. (2012), is that we estimate the within-…rm e¤ect of quality on export volumes. The

negative coe¢cient on quality indicates that when a …rm exports di¤erent wines with heterogeneous

levels of quality to a given destination in a given year, the higher quality wines are exported in smaller

quantities. This is consistent with San Martín, Troncoso, and Brümmer (2008) who observe that

higher quality wines are produced in smaller quantities, and with Thornton (2013) who notes that

higher quality wines are produced from grapes grown in low-yield vineyards. Besides, exports increase

with a real depreciation. The elasticity is large and equal to 1.378, which is consistent with evidence

that the trade elasticities for emerging economies are generally larger than for developed countries.47

Once we control for product-year …xed e¤ects in column (2), the exchange rate elasticity of export

volumes becomes insigni…cant. However, once we allow this elasticity to vary with the level of quality,

the results in column (3) show that, consistent with Prediction 2, the interaction between the exchange

rate and quality is negative and signi…cant, therefore the response of export volumes to changes in

exchange rates decreases with quality. This …nding remains robust to interacting the exchange rate with

year dummies (column 4), to including …rm-destination-year …xed e¤ects (column 5), to a di¤erence-

in-di¤erence speci…cation (column 6), and (less so) to di¤erent sample periods (columns 7 to 9).

46Given that markups may vary with destination market size (Melitz and Ottaviano, 2008), we checked, and con…rm,that our results remain robust to controlling for real GDP or real GDP per capita in our regressions.

47Haltmaier (2011) estimates an elasticity of total Argentinean exports to the real trade-weighted exchange rate of six.

18

In order to provide a quantitative evaluation of the economic e¤ects of quality, we discuss the

exchange rate elasticities evaluated at the mean value of quality in the sample (which is equal to 85),

as well as the elasticities calculated at the mid score of each quality bin de…ned in Table 3.48 In column

(3) of Panel A, the mean elasticity is equal to 0.193, therefore average pass-through is 80.7 percent.49

The elasticity is equal to 0.157 for “Not recommended” wines (at a score of 62), and increases to 0.180

for “Mediocre” wines (at a score of 77), which represents a 15 percent increase in pricing-to-market.

Moving from one quality bin to the other further increases pricing-to-market by an additional four

percent for “Good” (at a score of 82 and an elasticity of 0.188), “Very good” (at a score of 87 and an

elasticity of 0.196), “Outstanding” (at a score of 92 and an elasticity of 0.204), and “Great” wines (at a

score of 97.5 and an elasticity of 0.213). Pricing-to-market is thus about 36 percent higher for “Great”

than for “Not recommended” wines. The e¤ect of quality on pass-through is therefore economically

important, but smaller than in Auer et al. (2014) who …nd that the pass-through rate for a low quality

car is about four times larger than for a top quality car. In column (3) of Panel B, the exchange rate

elasticity of export volumes is equal to 0.858 at the mean value of quality. It is equal to 0.794 for

“Great” wines, and increases to 0.974 for “Not recommended” wines.

4.2 Heterogeneity across Destination Countries

We estimate equation (11) for unit values, and include interactions between the real exchange rate,

quality, and real GDP per capita. The results are reported in column (1) of Panel A in Table 11. At the

mean value of per capita income and of quality, the response of unit values to a change in the exchange

rate is equal to 0.150. Consistent with Prediction 1, this elasticity further increases with quality, and

is equal to 0.085 for “Not recommended” wines, and to 0.186 for “Great” wines. Most importantly,

the triple interaction between the exchange rate, quality, and real GDP per capita is positive and

signi…cant, which indicates that the exchange rate response of unit values to quality increases with per

capita income, providing direct support to Prediction 3. This …nding is consistent with the assumption

of Alessandria and Kaboski (2011) and Simonovska (2014) that consumers in richer countries have

smaller demand elasticities than consumers in lower income countries, allowing …rms to extract higher

markups from richer destinations.50 It is also consistent with the results of Alessandria and Kaboski

(2011) who show that US …rms price-to-market more when exporting to richer countries.

According to our model in Section 2, the exchange rate elasticity of unit values decreases with the

trade costs of exporting to each destination. We therefore check that our …ndings on the e¤ects of per

capita income are not a¤ected by di¤erences in trade costs between high and low income countries.

48The elasticities are evaluated at a value of 62 for the wines that are “Not recommended,” 77 for “Mediocre,” 82 for“Good,” 87 for “Very good,” 92 for “Outstanding,” and 97.5 for “Great” wines. The number of quality scores betweenbins is therefore not homogeneous, but comparing exchange rate elasticities across bins is intuitively convenient. Lookingat the e¤ect of a one standard deviation increase in quality from its mean level is not illustrative as this corresponds toless than four points on the quality scale, which is not large enough to move from one bin to the other.

49Strasser (2013) shows that …nancially constrained …rms price-to-market less than unconstrained …rms. In column (8),when restricting the sample to the period before the start of the …nancial crisis, the exchange rate elasticity evaluatedat the mean value of quality is equal to 0.332. Pricing-to-market is therefore stronger and pass-through is lower at 66.8percent, which is consistent with Argentinean exporters becoming more …nancially constrained during the crisis period.

50 It however contrasts with Crozet et al. (2012) who …nd that French exporters of Champagne charge higher prices tolower income countries. Their interpretation is that lower income countries do not produce much sparkling wine, thereforeFrench exporters charge higher prices when they enter the markets of these countries. We checked, and con…rm, that ourresults remain robust to controlling for the competition that may arise from wine production in destination countries.We estimated equations (11) and (12), and included total wine output (in tons) of each destination country in each year(Food and Agriculture Organization of the United Nations), and its interaction with the exchange rate.

19

Column (2) further interacts the exchange rate with the bilateral distance between Argentina and each

destination country.51 Consistent with the model, this interaction is negative and signi…cant. Most

importantly, controlling for bilateral distance does not alter the evidence in favor of Prediction 3 as the

triple interaction remains positive and signi…cant. Besides, at the mean value of quality, the exchange

rate elasticity of unit values is equal to 0.127. It is insigni…cant for “Not recommended” wines, but

increases to 0.163 for “Great” wines, again providing evidence in support of Prediction 1.

Also predicted by our model is that pricing-to-market increases with local distribution costs in

the importing country. As the latter can be correlated with per capita income, we verify that our

results are not driven by cross-country di¤erences in distribution costs. One way to proxy for local

distribution costs is to follow Goldberg and Verboven (2001) and to rely on the real wage in each

destination market as a proxy for local costs, under the assumption that labor costs represent a large

share of local distribution costs (also see Section 6.2). The data are not available for all countries so our

sample size is slightly reduced. We interact the exchange rate with the log of real wages, and the results

in column (3) show that, consistent with expectations, the response of unit values to a change in the

exchange rate increases with wages, and thus with distribution costs. This exercise does not, however,

qualitatively a¤ect the evidence in favor of Prediction 3 (only the magnitude of pricing-to-market is

increased as the exchange rate elasticity evaluated at the mean value of quality is equal to 0.362, and

rises from 0.296 for “Not recommended” wines to 0.398 for “Great” wines).

As the Wine Spectator is a US-based rating, a …nal concern is that it may not be representative

of taste preferences for quality in other countries.52 In column (4), we therefore drop the US from the

sample. Finally, in column (5) we simultaneously drop the US and control for the e¤ects of distance and

of wages on the responsiveness of unit values to exchange rate changes. On average, the exclusion of the

US from the sample decreases exchange rate pass-through (at the mean value of quality, the exchange

rate elasticities are equal to 0.196 and 0.364 in columns 4 and 5, respectively), but the evidence in

favor of Prediction 3 largely holds as the triple interactions between the exchange rate, quality, and

per capita income remain positive and signi…cant.53

The results for export volumes are reported in Panel B. The triple interactions between the ex-

change rate, quality, and per capita income are all negative, which is consistent with Prediction 4, but

insigni…cant. Still, consistent with Prediction 3, the exchange rate elasticities decrease with quality.

For instance, in column (5), the mean elasticity of export volumes is equal to 2.040, and it increases

from 1.992 for “Great” wines to 2.129 for “Not recommended” wines.

51The model does not predict that the exchange rate elasticity of prices to trade costs varies with per capita income.As a result, we do not interact this elasticity with per capita income.

52For instance, US consumers have a preference for fruiter wines with less alcoholic content while Europeans prefer lessfruity wines with higher alcohol content (Artopoulos, Friel, and Hallak, 2011).

53Some recent papers argue that the demand for quality is not only determined by income, but also by the distribution ofincome within a country, and that richer households typically spend a larger share of their income on higher quality goods.Fajgelbaum et al. (2011) derive conditions under which richer, as well as more unequal countries, have a larger demandfor higher quality goods. Mitra and Trindade (2005) show that more unequal countries import more of higher qualitygoods. If the demand for quality is indeed stronger for more unequal countries, the response of unit values to changes inthe exchange rate should increase with quality by more for more unequal countries, while simultaneously controlling forincome per capita. We estimated equation (11) for unit values, and interacted both the exchange rate and quality withthe Gini coe¢cient of each destination country (World Bank World Development Indicators), while controlling for percapita income. We found that the exchange rate response of unit values to quality increases by more for more equal thanfor more unequal countries, where the distribution of income tends to be more equal in higher income countries.

20

Another way to investigate our predictions is to split the countries included in our sample between

high and low income, using the World Bank’s classi…cation based on GNI per capita in 2011.54 We

estimate equations (11) and (12) and interact the exchange rate and its interaction with quality with a

dummy for higher and a dummy for lower income countries. The results for unit values are reported in

column (1) of Panel A in Table 12. Interestingly, the coe¢cient on the interaction between the exchange

rate and quality is positive and signi…cant for higher income destinations only. This …nding is consistent

with Prediction 3, and with Auer et al. (2014) who predict that exchange rate pass-through decreases

with quality for higher income countries only. Qualitatively, the results remain robust to controlling

for bilateral distance (column 2), wages (column 3), excluding the US (column 4), and to all of them

simultaneously (column 5). Consistent with Alessandria and Kaboski (2011), pricing-to-market is also

stronger for exports to richer destinations. For instance, in column (5), at the mean value of quality the

exchange rate elasticity is insigni…cant for lower income countries. It is instead signi…cant and equal

to 0.302 for higher income countries, and increases from 0.252 for “Not recommended” wines to 0.329

for “Great” wines. In Panel B, the results for export volumes provide direct support for Prediction

4 as in all cases, the interactions between the exchange rate and quality are signi…cantly negative for

higher income countries only.

5 Endogeneity

One concern with our estimations is the potential endogeneity of quality in explaining unit values and

export volumes. First, although the Wine Spectator ratings are produced from blind tastings where the

“price is not taken into account in scoring,” the “tasters are told [...] the general type of wine (varietal

and/or region) and the vintage” year.55 Similarly for Parker, “neither the price nor the reputation

of the producer/grower a¤ect the rating in any manner,” but the “tastings are done in peer-group,

single-blind conditions (meaning that the same types of wines are tasted against each other, and the

producers names are not known).”56 As a result, we cannot rule out that the basic information that the

experts hold about each wine, in combination with their own subjective preferences, a¤ects in a way or

the other their scores, leading to an endogeneity bias which direction is, however, unclear.57 Second,

some papers question the reliability and usefulness of wine ratings as a measure of quality. Ashenfelter

and Quandt (1999) and Quandt (2007) argue that wine ratings fail to convey any useful information

about quality, and observe that the disagreements between experts are substantial. Hodgson (2008)

…nds that experts are also inconsistent as they do not assign the same scores to the same wines at

di¤erent tastings. Measurement error in the quality scores can thus generate a positive endogeneity

bias as wine producers are likely to set a higher price for a wine that is awarded a higher score because

they believe consumers are willing to pay more for higher ranked wines. To tackle the endogeneity of

quality, we use appropriate instruments.

The set of instruments we rely on to explain the variation in wine quality includes geographic and

weather-related factors. Indeed, the literature devoted to explaining the quality of wine highlights that

the amount of rainfall, and the average temperatures during the growing season, are strong determinants

54Low income countries have a GNI per capita of less than 4,035 US dollars.55See http://www.winespectator.com/display/show?id=about-our-tastings.56See https://www.erobertparker.com/info/legend.asp.57Psychologists argue that experts ratings tend to be subjective and may depend on the personal preferences of the

tasters. For instance, Parker is known to have a strong preference for fruit-‡avored wines and rewards them with a higherscore (Thornton, 2013).

21

of quality (Ashenfelter, 2008; Ramirez, 2008; Thornton, 2013). In the Southern hemisphere, the growing

period spans the period from September (in the year before the vintage year) to March. In order to

allow for the e¤ects of temperature and rainfall to be nonlinear throughout the growing season, we

consider as instruments the average temperature, and the total amount of rainfall, for each growing

province in each month between September and March (Ramirez, 2008).58 Besides, one particularity

of Argentina’s wine industry is the high altitude at which some of the growing regions are located, and

there are strong reasons to believe that altitude contributes to variations in quality because it reduces

the problems related to insects or grape diseases that a¤ect quality at a low altitude. We therefore use

the altitude of each province as an additional instrument for quality.59 The data on monthly average

temperatures (in degrees Celsius), total rainfall (in millimeters), and altitude (in meters) are from the

National Climatic Data Center of the US Department of Commerce.60 Gaps in the data are …lled

using online information, although missing information for some provinces and vintage years results in

a slightly reduced sample.61

Endogeneity could also arise due to measurement error in the real exchange rate, and in particular

in the Argentinean CPI. Since 2007, the credibility of the o¢cial CPI data has indeed been widely

questioned by the public. In fact, the IMF has issued a declaration of censure, and called on Argentina

to adopt remedial measures to address the quality of the o¢cial CPI data.62 Alternative data sources

have shown considerably higher in‡ation rates than the o¢cial data since 2007.63 Using online data from

supermarket chains, Cavallo (2013) constructs a CPI for Argentina, and shows that the corresponding

in‡ation rate is three times larger than the o¢cial estimate. To address measurement error, we therefore

use the CPI from Cavallo (2013) for 2008 and 2009 to update the o¢cial CPI series, and construct a

new real exchange rate that we use in our estimations.

In order to establish how the instrumentation a¤ects the way quality determines unit values and

export volumes, we …rst focus on speci…cations where the product-year …xed e¤ects are replaced by

…rm-year dummy variables, and product characteristics are controlled for by including grape, type,

vintage year, HS, and province of origin …xed e¤ects. As the instruments are only available over a

reduced sample, we replicate our benchmark OLS estimations reported in column (1) of Panels A and

B of Table 10. The results, reported in column (1) of Panels A and B of Table 13 for prices and

quantities, respectively, show that our main …ndings go through over the smaller sample.

Column (2) of Panel A regresses by Instrumental Variables (IV) unit values on the real exchange

rate and quality. The coe¢cient on quality is positive and signi…cant, but is smaller compared to the

OLS estimate in column (1). This positive bias suggests that wine tasters tend to assign higher scores

to more expensive wines, or that producers charge a higher price when their wines receive a higher

score. Column (3) further shows that our IV estimates are not much a¤ected by using the CPI data

from Cavallo (2013), and are thus robust to measurement error in the real exchange rate. Panel B looks

58The temperatures, rainfall, and altitude are measured at the following weather stations (province): CatamarcaAero (Catamarca), Paraná Aero (Entre Ríos), Santa Rosa Aero (La Pampa), La Rioja Aero (La Rioja), Mendoza Aero(Mendoza), Neuquén Aero (Neuquen), Bariloche Aero (Río Negro), Salta Aero (Salta), and San Juan Aero (San Juan).

59One concern is that our instruments could be correlated with unobserved productivity shocks. The inclusion of…rm-year or product-year …xed e¤ects controls for these shocks so the exclusion restriction of our instruments is satis…ed.

60This is available at http://www.ncdc.noaa.gov/IPS/mcdw/mcdw.html.61Online information is taken from www.tutiempo.net.62Further information can be found at http://www.imf.org/external/np/sec/pr/2013/pr1333.htm.63See http://www.economist.com/node/21548242.

22

at export volumes. In column (2), the instrumented e¤ect of quality on export volumes is negative and

signi…cant, and is in turn larger in magnitude compared to the OLS estimate in column (1). Again,

the results remain similar once we use the adjusted CPI data in column (3). For all regressions in

columns (2) and (3), the Kleibergen-Paap F statistic (equal to 93, with a critical value equal to 21,

Stock and Yogo, 2005) largely rejects the null of weak correlation between the excluded instruments

and the endogenous regressors.

Next, we regress unit values on the exchange rate, quality, and its interaction with the exchange

rate. Columns (4), (5) and (6) report the OLS and IV results, using either the o¢cial or the adjusted

CPI data, respectively. The set of instruments for quality, and for the interaction term, includes

the monthly temperatures, rainfall, and altitude variables, as well as each of the variables interacted

with the exchange rate.64 Interestingly, in columns (5) and (6) the coe¢cient on the interaction term

increases to 0.014 from a value of 0.002 in column (4). The e¤ect of quality on pass-through is therefore

much larger in the instrumented regressions.

Columns (7) to (9) report the OLS and two IV speci…cations once we include the product-year …xed

e¤ects. As before, the instrumentation in columns (8) and (9) in‡ates the coe¢cient on the interaction

term at a value of 0.015 compared to an OLS estimate of 0.001 in column (7). As a result, the OLS

and IV exchange rate elasticities are respectively equal to 0.201 and -0.089 (but insigni…cant) for wines

that are “Not recommended,” 0.223 and 0.143 for “Mediocre,” 0.231 and 0.220 for “Good,” 0.238 and

0.298 for “Very good,” 0.245 and 0.375 for “Outstanding,” and to 0.253 and 0.460 for “Great” wines.

In other words, the instrumentation decreases exchange rate pass-through for the higher quality wines.

In addition, notice that the di¤erence in pricing-to-market between “Great” and “Outstanding” wines

is equal to 23 percent once we instrument, and to three percent otherwise. For export volumes in Panel

B, the IV coe¢cients on the interaction terms are not statistically signi…cant.

The …rst-stage estimates of the IV regressions in column (5) of Panels A and B (which use the

o¢cial CPI data) are reported in Table A1 in the appendix (the …rst-stages for the other regressions

are qualitatively similar), and show that climate variation a¤ects wine quality. The positive coe¢cients

on the January and February temperatures are consistent with the …nding in the literature that warmer

temperatures during the harvest period are typically associated with higher quality. Also, the positive

coe¢cient on the October rainfall, and the negative coe¢cients on the January and February precip-

itations, are consistent with the expectation that precipitation during the earlier part of the growing

season is good for quality, while a dry climate during the harvest period is more favorable for crops

(Ramirez, 2008).

6 Extensions

This section discusses …ve extensions to our benchmark speci…cations. First, we allow for nonlinearities

in the e¤ects of quality. Second, we show that pricing-to-market increases with local distribution costs in

the importing economy. Third, we provide evidence that in addition to increasing with quality, pricing-

to-market also increases with …rm-level productivity. Fourth, once we divide the sample between high

64 In columns (2) and (3), which exclude the interaction between the exchange rate and quality, altitude is dropped as aninstrument as it is collinear with the province …xed e¤ects. In columns (5), (6), (8) and (9) which include the interactionterm, the interaction between the exchange rate and altitude is used as an instrument.

23

and low quality exporters, higher quality …rms price-to-market more than others. Finally, …rms price-

to-market more in the countries where they own a larger share of the market.

6.1 Nonlinearities

The prediction of our model that pass-through decreases with quality depends crucially on the assump-

tion that higher quality goods have higher (per unit) distribution costs, which include any wholesale

or retail costs associated with the selling, storage, marketing, advertising, or transport of each good in

the destination country.65 If (per unit) distribution costs are indeed higher for top quality wines, our

model predicts that the elasticity of prices to the exchange rate should be larger for top quality wines.

To check this assumption, we estimate equation (11) for unit values, allowing for nonlinearities in the

e¤ects of quality by using six variables for quality, i.e., one for each of the six quality bins de…ned in

Table 3, each interacted with the exchange rate.

The results are reported in Table 14. Column (1) shows that the interaction between the exchange

rate and quality is about three times larger for “Great” than for lower ranked wines. As we cannot

reject that the coe¢cients on the interactions for the lower quality wines are equal, in column (2) we

classify wines into two di¤erent bins only, i.e., with a score above or below 94. The coe¢cient on the

interaction is still much larger for “Great” wines, a …nding which can be reconciled with our assumption

that higher quality goods have higher (per unit) distribution costs.

In column (2), when calculating the exchange rate elasticities evaluated at the mid score of each

quality bin, we …nd that in response to a ten percent real depreciation, exporters raise their prices (in

pesos) by 1.8 percent for the lowest quality wines (at a mid score of 72), and by 6.5 percent for the

top quality wines (at a mid score of 97.5). This di¤erence is thus much larger than in our benchmark

speci…cation, reported in column 3 of Table 10, where pricing-to-market is 36 percent higher for “Great”

than for “Not recommended” wines. The results for export volumes are reported in columns (3) and (4).

In column (4), the coe¢cient on the interaction between the exchange rate and quality is signi…cantly

negative for the lower ranked wines only.

6.2 Distribution Costs

Our model predicts that pricing-to-market increases with local distribution costs in the importing

economy. One way to explore this prediction is to use the data on distribution costs collected by

Campa and Goldberg (2010) for 21 countries and 29 industries between 1995 and 2002, and to compute

a destination-speci…c measure of distribution costs for the “Food products and beverages” industry

averaged over time. However, due to limited country coverage, this approach reduces our sample size

by more than half. As in Section 4.2, we therefore instead follow Goldberg and Verboven (2001) and

rely on the real wage in each destination market as a proxy for local costs.66

We divide our sample into two groups, namely, high wage and low wage destination countries,

according to whether their wage is above or below the median wage in the sample. Columns (1) and

(2) of Table 15 report the results for unit values and export volumes, separately for countries with high

65Using models that do not feature this assumption, Antoniades and Zaniboni (2013), Auer and Chaney (2009), Aueret al. (2014), and Basile et al. (2012) also predict a negative relationship between pass-through and quality.

66Although higher quality wines can be expected to have higher distribution costs, in this section we focus on di¤erencesin local distribution costs across export destinations.

24

and low wages. Consistent with expectations, Panel A shows that …rms price-to-market more, and by

more for higher quality wines, in the countries with higher wages, and therefore higher distribution

costs. At the mean value of quality, the exchange rate elasticity is signi…cant at 0.206 in higher wage

countries, but insigni…cant in lower wage countries. This …nding is consistent with Berman et al.

(2012) who show that the response of unit values to changes in the exchange rate increases with local

costs, especially for higher productivity …rms. Symmetrically, Panel B shows that in response to a

real depreciation, exporters increase export volumes only when exporting to lower distribution costs

countries (at the mean value of quality, the elasticity is insigni…cant for higher wage countries, but

signi…cant at a value of 2.843 for lower wage countries).

6.3 Productivity as a Source of Heterogeneity

As the literature has focused on productivity in explaining incomplete and heterogeneous pass-through

(Berman et al., 2012; Chatterjee et al., 2013), we explore whether, in our data, higher productivity

…rms also price-to-market more than the others.

Without any information on the …rm-level characteristics that are required to measure productivity

(such as value-added or employment), we construct a variable that proxies for …rm size as the latter

has been shown to correlate strongly with productivity (see, for instance, Bernard, Eaton, Jensen, and

Kortum, 2003).67 We use the total volume (in liters) of FOB exports of each …rm in each year, and

divide our sample into two groups of larger and smaller …rms, according to whether their total yearly

exports are above or below the median total exports in the sample. Columns (3) and (4) of Table 15

report the results for unit values and export volumes, separately for larger and smaller …rms. Panel

A shows that only the larger, and therefore the more productive …rms, signi…cantly price-to-market in

response to a change in the exchange rate (at the mean value of quality, the elasticity is signi…cant at

0.164 in column 3, but insigni…cant in column 4), and they do so by more for higher quality wines.

Symmetrically, Panel B reveals that smaller …rms increase their export volumes by more than the others

(at the mean value of quality, their exchange rate elasticity is signi…cant at 2.762, but the elasticity is

insigni…cant for larger …rms). However, the response of export volumes to a real depreciation decreases

with quality for larger …rms only.

6.4 The Quality of Exporters

As illustrated by Table 7, individual …rms tend to export wines with heterogeneous levels of quality.

Table 9 shows that there is also some heterogeneity in the average quality exported by …rms. Firms

can therefore be divided into two groups of higher and lower quality exporters by splitting the sample

at the median of average …rm-level quality, which varies between 72 and 92. This sample split is not

perfect as average …rm-level quality does not include unrated wines, but it still captures that …rms

export wines with heterogeneous levels of quality: quality varies between 55 and 97 for higher quality

exporters (average quality is equal to 87), and between 55 and 94 for lower quality exporters (average

quality is equal to 83).

Columns (1) and (2) of Table 16 report the results for unit values and export volumes, separately

for higher and lower quality exporters. At the mean value of quality, pricing-to-market is stronger

67 In the case of wine, larger …rms can adopt productivity enhancing equipment and technologies that cannot be usedby smaller …rms (Thornton, 2013).

25

for higher than for lower quality exporters (the elasticities are equal to 0.258 and 0.127, respectively).

Symmetrically, lower quality exporters increase export volumes by more than higher quality …rms do

(with an elasticity of 0.957 versus 0.848 for higher quality …rms). Pricing-to-market increases with

quality for higher quality exporters only, while the response of export volumes to a change in the

exchange rate decreases with quality for both higher and lower quality exporters.

6.5 Export Market Shares

According to Amiti et al. (2014), exporters have smaller markups in the markets where they have a

lower market share, making it di¢cult to adjust their markups in response to changes in the exchange

rate. As a result, …rms price-to-market less, and have a higher pass-through, in the countries where

they own a smaller share of the market (see, also, Atkeson and Burstein, 2008). To check if this holds

in our data, we compute the market share of each …rm by export destination and by year as the share

of its total exports in the total value exported by all …rms. Based on the median market share in each

destination and in each year, we then divide …rms into two groups, namely, high and low market share

…rms. The distribution of quality is not much a¤ected by this split as quality varies between 55 and

97 for higher market share …rms, and between 55 and 96 for lower market share …rms.

Columns (3) and (4) of Table 16 report the results for unit values and export volumes, separately

for exporters with higher and lower market shares. Consistent with Amiti et al. (2014), Panel A

shows that, at the mean value of quality, pass-through is incomplete at 81 percent for exporters with

higher market shares (the elasticity is equal to 0.190), and is complete for lower market share exporters

(the elasticity is insigni…cant). Pass-through decreases with quality only for the exporters with higher

market shares.

The results for volumes are reported in Panel B. Columns (3) and (4) show that a real depreciation

increases the export volumes of the higher market share exporters by less for higher quality goods, but

has no e¤ect on the export volumes of the lower market share exporters, at any level of quality. In other

words, a real depreciation only bene…ts the higher market share exporters. One possible interpretation

is that market shares signal quality. In response to a real depreciation, the higher market share

exporters expand at the expense of the others because consumers in the destination country believe

that these …rms produce higher quality goods. Caminal and Vives (1996) develop a theory of the

informational value of market shares, and show that consumers with private but imperfect information

on quality trust that a higher market share signals a higher quality. Another related interpretation

is that in contrast to the lower market share …rms, the higher market share exporters bene…t from a

real depreciation because they have acquired a good reputation in terms of quality and reliability in

the destination market. Macchiavello (2010) provides evidence that reputation is indeed an important

determinant of a …rm’s success in export markets.68

68To summarize, our results show that larger, and therefore more productive …rms, exporting a higher quality andowning higher shares of export markets price-to-market more than the others, and by more for higher quality goods.Larger …rms tend to have higher market shares (the correlation is equal to 0.53). In contrast, the higher quality …rms arenot larger, and do not own higher export market shares (the correlations between …rm-level quality, on the one hand, andmarket shares and …rm size, on the other hand, are equal to -0.06 and -0.04, respectively).

26

7 Robustness

This section discusses alternative speci…cations we implement to ensure the robustness of our …ndings.

Despite some variation across speci…cations in the magnitude of the e¤ect of quality on pass-through,

the broad similarity of the results supports the paper’s main conclusions.

The Measurement of Quality In order to minimize noise in the measurement of quality when

de…ned on a (50,100) scale, in column (1) of Table 17 we use a variable which takes on values between

one and six, where each value corresponds to one of the di¤erent bins de…ned by the Wine Spectator

(see Table 3). A value of one indicates that the wine is “Not recommended,” while a value of six that

the wine is “Great.” In column (2), quality is measured using the Parker ratings. In column (3), in

order to reduce the subjectiveness of the quality scores, and the fact that di¤erent tasters disagree on

wine quality, we only include the wines which Wine Spectator scores fall in the same quality bin as

the Parker ratings (as shown in Figure 2, these wines belong to categories four, …ve, and six only).

Finally, given that both the Wine Spectator and Parker are US-based ratings, and therefore may not

be representative of perceived wine quality in other countries, column (4) excludes the US from the

sample. Qualitatively, our results largely hold up.

Most of the empirical literature on quality and trade relies on trade unit values as a proxy for quality.

Indeed, our regressions indicate that on average, unit values increase with quality. We therefore check

if our results remain robust to measuring quality using the information contained in unit values. One

advantage of this approach is that we can estimate our regressions on a much larger sample, including

all wines for which the Wine Spectator or the Parker ratings are missing.

First, in each year, we calculate the mean unit value (in pesos) of each wine across all destinations,

and for each …rm we rank its wines by unit values in decreasing order (Berman et al., 2012; Chatterjee

et al., 2013; Mayer et al., 2014). For each …rm, the wine with the highest average unit value has a rank

equal to one, the second a rank equal to two, etc., and we use these ranks as an inverted measure for

quality. Second, we repeat the procedure, but for each wine exported to all destinations in all years,

using the mean unit value of each wine (in pesos) de‡ated by the Argentinean CPI. The interactions

between the exchange rate and the product ranks are expected to be negative for unit values, and

positive for export volumes.

The results, reported in columns (5) and (6), are consistent with expectations, and therefore contrast

with Auer and Chaney (2009) who do not …nd any evidence that pass-through is a¤ected by quality

when quality is inferred from trade unit values. Columns (5) and (6) use the product ranks computed

by …rm-product-year, and by …rm-product, respectively.

Unrated Wines Recall that due to missing observations on the Wine Spectator ratings, our sample

covers 43 percent of the total value of red, white, and rosé wine exported by Argentina between 2002 and

2009. In order to include some unrated wines in the sample, we calculate an average Wine Spectator

rating by wine name and type, and assign this rating to all wines with the same name and type. This

increases our sample coverage to 63 percent of total exports over the period. We apply this procedure

to compute average quality both on a (50,100) and on a (1,6) scale. The results are reported in columns

(7) and (8) of Table 17.

27

Another way to include unrated wines in the sample is as follows. First, we identify the wines which

vintage year is missing, and assign to each of them the scores of the wines with the same name, brand,

and type on a (1,6) scale. In general, quality does not vary much across vintage years, therefore this

assumption sounds reasonable.69 Second, we assign a value of one to the wines exported by …rms that

only export unrated wines (by either the Wine Spectator or Parker). We do this under the assumption

that these …rms produce wines which are of a too low quality to be considered by experts (this approach

is similar to Crozet et al., 2012, who assume that unrated …rms are the lowest quality exporters). This

exercise increases our sample coverage to 60 percent of total exports between 2002 and 2009. The

results are reported in column (9) of Table 17 (the interaction between the exchange rate and quality

is insigni…cant for export volumes).

The Nature of Wine Our results might be a¤ected by two characteristics that are speci…c to the

nature of wine. First, wine is an exhaustible resource: once a wine with a speci…c vintage year runs

out, the producer can no longer produce, and therefore export, that variety. Second, the production

of higher quality wines is subject to capacity constraints as the availability of high quality grapes is

limited (San Martín et al., 2008; Thornton, 2013).

In order to control for the exhaustible nature of wine, we construct a new sample and de…ne a

product according to the name of the wine, its grape, and type, but ignore the vintage year.70 The

209 …rms in the sample export a smaller number of products (2,790 versus 6,720 wines in the original

sample) to a larger number of countries (78 versus 24 destinations, on average). Quality is computed

as the unweighted average of the scores assigned to all wines with the same name, type, and grape, and

varies between 55 and 96.2.71 The correlation between average quality (across vintage years) and the

original quality measure (which varies across vintage years) is equal to 93.7 percent, con…rming that

quality does not vary much across vintage years. The results for unit values and volumes are reported

in columns (1) and (3) of Table 18.

To address the issue of capacity constraints, we would need to control for total output per wine

which is unobserved. As a proxy, we rely on the total volume (in liters) of exports of each wine to

all destinations between 2002 and 2009. Total exports are however endogenous to the denominator

of unit values in equation (11), and to the dependent variable in equation (12). We therefore classify

wines into three categories (based on the 33 and 66 percentiles of total exports), and create an

indicator variable which varies between one and three, with a larger value indicating a greater volume

of exports.72 We then include this indicator, on its own and interacted with the exchange rate, in

columns (2) and (4) of Table 18 for prices and quantities, respectively. Our results remain unaltered.

Exchange Rate After the large devaluation of the peso in 2002, the peso was allowed to ‡uctuate

with respect to the US dollar within a “crawling band that is narrower than or equal to +/-2 percent”

(Reinhart and Rogo¤, 2004). This means that variations in the real exchange rate between the peso

69 If wine producers conclude that the grapes grown during a particular year do not satisfy their quality standards, theymay decide not to use them in order to preserve the quality and reputation of a wine (Thornton, 2013).

70This exercise allows us to get closer to the CES utility function assumption of a …xed set of varieties as there is noreason for wines to run out once we exclude the vintage year.

71The results remain similar if quality is computed as a weighted average of the original quality scores, with weightsgiven by the export shares of each wine to each destination country in each year.

72The results remain similar if we use the total volume of exports instead of the indicator variable.

28

and the US dollar may have essentially come from movements in domestic prices (Li et al., 2012). We

verify in column (1) of Table 19 that our results still hold after excluding from the sample the US and

all the other countries which currencies are pegged to the US dollar.

In order to address the quality of the o¢cial CPI data, in column (2) of Table 19 we use the real

exchange rate computed using the CPI from Cavallo (2013).73 Finally, although the model in Section

2 derives predictions for the e¤ects of the real exchange rate, many papers estimate pass-through

regressions using the nominal rate (e.g., Gopinath et al., 2010; Gopinath and Rigobon, 2008). Column

(3) of Table 19 includes the nominal exchange rate, its interaction with quality, and separately controls

for the destination country’s CPI.

Extensive Margin Campos (2010) argues that the intensive and extensive margins of adjustment

have opposite e¤ects on pass-through. On the one hand, a depreciation reduces the average price

charged by existing exporters. On the other hand, a depreciation makes exporting a more pro…table

activity, therefore more …rms enter the export market. Given that entrants are less productive and

therefore charge higher prices, the extensive margin pushes the average export price up, reducing pass-

through. In column (1) of Table 20, we estimate equations (11) and (12) on a sample that captures

the intensive margin, and only includes the …rms exporting in all years to any destination.

Quarterly Frequency We check in column (2) of Table 20 that our results remain robust to higher

frequency sampling, and estimate equations (11) and (12) using unit values, export volumes, and

the real exchange rate de…ned at a quarterly frequency, and replace the product-year …xed e¤ects by

product-quarter dummy variables. Due to data limitations, the real GDPs per capita and real e¤ective

exchange rates are measured annually.

Dynamics In order to introduce dynamics in equations (11) and (12), we include one lag, and then

two lags, on the exchange rate and its interaction with quality. In the volumes regressions, lags on

the real e¤ective exchange rate and real GDP per capita are also included. The results are reported

in columns (3) and (4) of Table 20 for the speci…cations with one and two lags, respectively. In both

columns, the estimated coe¢cients are the sum of the coe¢cients on the contemporaneous values and

the lags of each variable (Gopinath and Itskhoki, 2010; Gopinath et al., 2010).

Wholesalers and Retailers We have restricted our analysis to wine producers only, and therefore

dropped wholesalers and retailers from the sample. As can be seen from column (5) of Table 20, our

results still hold when including these …rms in the sample.

Currency of Invoicing A large body of the recent literature is devoted to understanding how the

currency of invoicing used for trade a¤ects exchange rate pass-through (e.g., Gopinath et al., 2010).

In our data set, we do not observe the currency in which Argentinean …rms price their exports. The

Datamyne, a private vendor of international trade data, provides us with the invoicing currency of

…rm-level exports between 2005 and 2008. Over the period, Argentinean …rms priced wine exports

mostly in US dollars (88 percent), followed by euros (7.6 percent), Canadian dollars (3 percent), pound

sterling (1.2 percent), and in some cases in Japanese yen, Swiss francs, Uruguayan pesos, Australian

73Given that we adjust the CPI data for two years only, the results in column (2) remain very similar to the ones incolumn (3) of Table 10. We checked, and con…rm, that all our other results are robust to using the Cavallo (2013) index.

29

dollars, or Danish krones. Due to the predominance of the US dollar as an invoicing currency for

exports, the regression in column (6) of Table 20 expresses unit values in US dollars per liter.

8 Concluding Remarks

This paper analyzes the heterogeneous behavior of exporters to changes in real exchange rates due to

di¤erences in product quality. In order to understand the mechanisms through which quality a¤ects

the pricing and quantity decisions of …rms, the …rst contribution of the paper is to present a model

where …rms export multiple products with heterogeneous levels of quality. The second contribution is

to bring the testable predictions of the model to the data. We combine a unique data set of Argentinean

…rm-level destination-speci…c export values and volumes of highly disaggregated wine products between

2002 and 2009 with a data set on experts wine ratings to measure quality. Given the very rich nature of

our data set, our main sample includes 6,720 di¤erent wines exported by 209 wine producers over the

period. Overall, our results …nd strong support for the predictions of the model, and quantitatively,

the e¤ect of quality on heterogeneous pass-through is economically important.

To conclude, our …ndings help to explain incomplete pass-through by highlighting the role played

by product quality. As we are only focusing on a single industry, we do not know whether the empirical

regularities documented in this paper hold more generally, although the results of Auer et al. (2014) for

cars, and of Antoniades and Zaniboni (2013) for consumer goods, are consistent with ours. Provided

better data to measure quality for other industries become available, future research should test whether

our results extend beyond the Argentinean wine industry. Another promising avenue for future research

is to use our data to identify the pricing-to-market behavior of exporters by assuming that di¤erences

in the prices of the same wines sold to multiple destinations re‡ect di¤erences in markups as these

wines are subject to the same marginal cost. This is the approach adopted by Auer et al. (2014),

Burstein and Jaimovich (2012), Cavallo et al. (2014), Fitzgerald and Haller (2014), and Simonovska

(2014).

References

Alessandria, George, and Joseph P. Kaboski. 2011. “Pricing-to-Market and the Failure of Absolute

PPP.” American Economic Journal: Macroeconomics 3 (1): 91–127.

Amiti, Mary, Oleg Itskhoki, and Jozef Konings. 2014. “Importers, Exporters, and Exchange Rate

Disconnect.” American Economics Review 104 (7): 1942–78.

Amiti, Mary, and Amit K. Khandelwal. 2013. “Import Competition and Quality Upgrading.” Review

of Economics and Statistics 95 (2): 476–90.

Antoniades, Alexis, and Nicola Zaniboni. 2013. “Retailer Pass-Through and its Determinants Using

Scanner Data.” http://explore.georgetown.edu/people/aa658/.

Atkeson, Andrew, and Ariel Burstein. 2008. “Trade Costs, Pricing-to-Market, and International Rela-

tive Prices.” American Economic Review 98 (5): 1998–2031.

Auer, Raphael A., and Thomas Chaney. 2009. “Exchange Rate Pass-Through in a Competitive Model

of Pricing-To-Market.” Journal of Money, Credit and Banking 41 (1): 151–75.

Auer, Raphael A., Thomas Chaney, and Philip Sauré. 2014. “Quality Pricing-To-Market.”

https://sites.google.com/site/raphaelauer/.

30

Ashenfelter, Orley, and Richard Quandt. 1999. “Analyzing a Wine Tasting Statistically.” Chance 12

(3): 16–20.

Ashenfelter, Orley. 2008. “Predicting the Quality and Prices of Bordeaux Wine.” Economic Journal

118 (529): F174–F184.

Baldwin, Richard E., and James Harrigan. 2011. “Zeros, Quality, and Space: Trade Theory and Trade

Evidence.” American Economic Journal: Microeconomics 3 (2): 60–88.

Basile, Roberto, Sergio de Nardis, and Alessandro Girardi. 2012. “Pricing To Market, Firm Hetero-

geneity and the Role of Quality.” Review of World Economics 148 (4): 595–615.

Bastos, Paulo, and Joana Silva. 2010. “The Quality of a Firm’s Exports: Where You Export To

Matters.” Journal of International Economics 82 (2): 99–111.

Bastos, Paulo, Joana Silva, and Eric Verhoogen. 2014. “Export Destinations and Input Prices.” NBER

Working Paper 20143.

Berman, Nicolas, Philippe Martin, and Thierry Mayer. 2012. “How Do Di¤erent Exporters React to

Exchange Rate Changes?” Quarterly Journal of Economics 127 (1): 437–92.

Bernard, Andrew B., Jonathan Eaton, J. Bradford Jensen, and Samuel Kortum. 2003. “Plants and

Productivity in International Trade.” American Economic Review 93 (4): 1268–90.

Berthou, Antoine, and Charlotte Emlinger. 2011. “The Trade Unit Values Database.” CEPII Working

Paper 2011-10.

Burstein, Ariel, Martin Eichenbaum, and Sergio Rebelo. 2005. “Large Devaluations and the Real

Exchange Rate.” Journal of Political Economy 113 (4): 742–84.

Burstein, Ariel, and Nir Jaimovich. 2012. “Understanding Movements in Aggregate and Product-Level

Real Exchange Rates.” http://www.econ.ucla.edu/arielb/productRER.pdf.

Burstein, Ariel, and Gita Gopinath. 2014. “International Prices and Exchange Rates.” In Handbook of

International Economics 4: 391–451. Elsevier.

Burstein, Ariel, João C. Neves, and Sergio Rebelo. 2003. “Distribution Costs and Real Exchange Rate

Dynamics During Exchange-Rate-Based Stabilizations.” Journal of Monetary Economics 50 (6): 1189–

1214.

Caminal, Ramon, and Xavier Vives. 1996. “Why Market Shares Matter: An Information-Based The-

ory.” RAND Journal of Economics 27 (2): 221–39.

Campa, José M., and Linda S. Goldberg. 2005. “Exchange-Rate Pass-Through into Import Prices.”

Review of Economics and Statistics 87 (4): 679–90.

Campa, José M., and Linda S. Goldberg. 2010. “The Sensitivity of the CPI to Exchange Rates: Dis-

tribution Margins, Imported Inputs, and Trade Exposure.” Review of Economics and Statistics 92 (2):

392–407.

Campos, Camila F.S. 2010. “Incomplete Exchange Rate Pass-Through and Extensive Margin of Ad-

justment.” http://old-hha.asb.dk/nat/workshop/2010/cc1801.pdf.

Cavallo, Alberto. 2013. “Online and O¢cial Price Indexes: Measuring Argentina’s In‡ation.” Journal

of Monetary Economics 60 (2): 152–65.

Cavallo, Alberto, Brent Neiman, and Roberto Rigobon. 2014. “Currency Unions, Product Introduc-

tions, and the Real Exchange Rate.” Quarterly Journal of Economics 129 (2): 529–95.

Chatterjee, Arpita, Rafael Dix-Carneiro, and Jade Vichyanond. 2013. “Multi-Product Firms and Ex-

change Rate Fluctuations.” American Economic Journal: Economic Policy 5 (2): 77–110.

Chen, Natalie, and Luciana Juvenal. 2015. “Quality and the Great Trade Collapse.”

31

Corsetti, Giancarlo, and Luca Dedola. 2005. “A Macroeconomic Model of International Price Discrim-

ination.” Journal of International Economics 67 (1): 129–55.

Crinò, Rosario, and Paolo Epifani. 2012. “Productivity, Quality, and Export Behaviour.” Economic

Journal 122 (565): 1206–43.

Crozet, Matthieu, Keith Head, and Thierry Mayer. 2012. “Quality Sorting and Trade: Firm-Level

Evidence for French Wine.” Review of Economic Studies 79 (2): 609–44.

Dixit, Avinash K., and Joseph E. Stiglitz. 1977. “Monopolistic Competition and Optimum Product

Diversity.” American Economic Review 67 (3): 297–308.

Dornbusch, Rudiger. 1989. “Real Exchange Rates and Macroeconomics: A Selective Survey.” Scandi-

navian Journal of Economics 91 (2): 401–32.

Engel, Charles. 2003. “Expenditure Switching and Exchange Rate Policy.” In NBER Macroeconomics

Annual 17: 231-72. Chicago: University of Chicago Press.

Fajgelbaum, Pablo, Gene M. Grossman, and Elhanan Helpman. 2011. “Income Distribution, Product

Quality, and International Trade.” Journal of Political Economy 119 (4): 721–65.

Fan, Haichao, Yao Amber Li, and Stephen R. Yeaple. 2014. “Trade Liberalization, Quality, and Export

Prices.” NBER Working Paper 20323.

Feenstra, Robert C., and John Romalis. 2014. “International Prices and Endogenous Quality.” Quar-

terly Journal of Economics 129 (2): 477–527.

Fitzgerald, Doireann, and Stefanie Haller. 2014. “Pricing-to-Market: Evidence From Plant-Level

Prices.” Review of Economic Studies 81 (2): 761–86.

Fosse, Henrik B. 2012. “Exporter Price Response to Exchange Rate Changes.”

http://openarchive.cbs.dk/handle/10398/8539.

Gagnon, Etienne, Benjamin R. Mandel, and Robert J. Vigfusson. 2014. “Missing Import Price Changes

and Low Exchange Rate Pass-Through.” American Economic Journal: Macroeconomics 6 (2): 156–

206.

Goldberg, Pinelopi K., and Michael M. Knetter. 1997. “Goods Prices and Exchange Rates: What Have

We Learned?” Journal of Economic Literature 35 (3): 1243–72.

Goldberg, Pinelopi K., and Frank Verboven. 2001. “The Evolution of Price Dispersion in the European

Car Market.” Review of Economic Studies 68 (4): 811–48.

Gopinath, Gita, Oleg Itskhoki, and Roberto Rigobon. 2010. “Currency Choice and Exchange Rate

Pass-Through.” American Economic Review 100 (1): 304–36.

Gopinath, Gita, and Oleg Itskhoki. 2010. “Frequency of Price Adjustment and Pass-Through.” Quar-

terly Journal of Economics 125 (2): 675–727.

Gopinath, Gita, and Roberto Rigobon. 2008. “Sticky Borders.” Quarterly Journal of Economics 123

(2): 531–75.

Hallak, Juan Carlos. 2006. “Product Quality and the Direction of Trade.” Journal of International

Economics 68 (1): 238–65.

Hallak, Juan Carlos, and Peter K. Schott. 2011. “Estimating Cross-Country Di¤erences in Product

Quality.” Quarterly Journal of Economics 126 (1): 417–74.

Hallak, Juan Carlos, and Jagadeesh Sivadasan. 2011. “Firms’ Exporting Behavior under Quality Con-

straints.” Research Seminar in International Economics, The University of Michigan, Discussion Paper

628.

32

Haltmaier, Jane. 2011. “Empirical Estimation of Trend and Cyclical Export Elasticities.” Board of

Governors of the Federal System, International Finance Discussion Paper 1030.

Hellerstein, Rebecca. 2008. “Who Bears the Cost of a Change in the Exchange Rate? Pass-Through

Accounting in the Case of Beer.” Journal of International Economics 76 (1): 14–32.

Hodgson, Robert T. 2008. “An Examination of Judge Reliability at a Major US Wine Competition.”

Journal of Wine Economics 3 (2): 105–13.

Hummels, David, and Peter J. Klenow. 2005. “The Variety and Quality of a Nation’s Exports.” Amer-

ican Economic Review 95 (3): 704–23.

Hummels, David, and Alexandre Skiba. 2004. “Shipping the Goods Apples Out? An Empirical Con-

…rmation of the Alchian-Allen Conjecture.” Journal of Political Economy 112 (6): 1384–1402.

Jaimovich, Esteban, and Vincenzo Merella. 2014. “Love for Quality, Comparative Advantage, and

Trade.” https://sites.google.com/site/vincenzomerella/research.

Johnson, Robert C. 2012. “Trade and Prices with Heterogeneous Firms.” Journal of International

Economics 86 (1): 43–56.

Joseph, Robert. 2012. “Lifting the Stone on the UK Wine Trade.”

http://thejosephreport.blogspot.fr/2012/08/lifting-stone-on-uk-wine-trade.html.

Kasa, Kenneth. 1992. “Adjustment Costs and Pricing to Market.” Journal of International Economics

32 (1/2): 1–30.

Khandelwal, Amit. 2010. “The Long and Short (of) Quality Ladders.” Review of Economic Studies 77

(4): 1450–76.

Knetter, Michael M. 1989. “Price Discrimination by US and German Exporters.” American Economic

Review 79 (1): 198–210.

Knetter, Michael M. 1992. “Exchange Rates and Corporate Pricing Strategies.” NBER Working Paper

4151.

Knetter, Michael M. 1993. “International Comparisons of Pricing-To-Market Behavior.” American

Economic Review 83 (3): 473–86.

Kugler, Maurice, and Eric Verhoogen. 2012. “Prices, Plant Size, and Product Quality.” Review of

Economic Studies 79 (1): 307–39.

Li, Hongbin, Hong Ma, Yuan Xu, and Yanyan Xiong. 2012. “How Do Exchange Rate Movements A¤ect

Chinese Exports? A Firm-Level Investigation.” http://mahong.weebly.com/research.html.

Macchiavello, Rocco. 2010. “Development Uncorked: Reputation Acquisition in the New Market for

Chilean Wines in the UK.” http://www2.warwick.ac.uk/fac/soc/economics/sta¤/rmacchiavello.

Manova, Kalina, and Zhiwei Zhang. 2012a. “Export Prices Across Firms and Destinations.” Quarterly

Journal of Economics 127 (1): 379–436.

Manova, Kalina, and Zhiwei Zhang. 2012b. “Multi-Product Firms and Product Quality.” NBER Work-

ing Paper 18637.

Marston, Richard C. 1990. “Pricing To Market in Japanese Manufacturing.” Journal of International

Economics 29 (3/4): 217–36.

Mayer, Thierry, Marc J. Melitz, and Gianmarco I.P. Ottaviano. 2014. “Market Size, Competition, and

the Product Mix of Exporters.” American Economic Review 104 (2): 495–536.

Melitz, Marc J., and Gianmarco I.P. Ottaviano. 2008. “Market Size, Trade, and Productivity.” Review

of Economic Studies 75 (1): 295–316.

33

Mitra, Devashish, and Vitor Trindade. 2005. “Inequality and Trade.” Canadian Journal of Economics

38 (4): 1253–71.

Nakamura, Emi, and Jón Steinsson. 2012. “Lost in Transit: Product Replacement Bias and Pricing to

Market.” American Economic Review 102 (7): 3277–3316.

Nakamura, Emi, and Dawit Zerom. 2010. “Accounting for Incomplete Pass-Through.” Review of Eco-

nomic Studies 77 (3): 1192–1230.

Quandt, Richard E. 2007. “On Wine Bullshit: Some New Software?” Journal of Wine Economics 2

(2): 129–35.

Ramirez, Carlos D. 2008. “Wine Quality, Wine Prices, and the Weather: Is Napa Di¤erent?” Journal

of Wine Economics 3 (2): 114–31.

Reinhart, Carmen M., and Kenneth S. Rogo¤. 2004. “The Modern History of Exchange Rate Arrange-

ments: A Reinterpretation.” Quarterly Journal of Economics 119 (1): 1–48.

Rodríguez-López, José A. 2011. “Prices and Exchange Rates: A Theory of Disconnect.” Review of

Economic Studies 78 (3): 1135–77.

San Martín, Guillermo J., Javier L. Troncoso, and Bernhard Brümmer. 2008. “Determinants of Argen-

tinean Wine Prices in the US.” Journal of Wine Economics 3 (1): 72–84.

Simonovska, Ina. 2014. “Income Di¤erences and Prices of Tradables: Insights from an Online Retailer.”

http://inasimonovska.weebly.com/uploads/1/3/2/5/1325220/mango_2014.pdf.

Schott, Peter K. 2004. “Across-Product versus Within-Product Specialization in International Trade.”

Quarterly Journal of Economics 119 (2): 647–78.

Strasser, Georg. 2013. “Exchange Rate Pass-Through and Credit Constraints.” Journal of Monetary

Economics 60 (1): 25–38.

Stock, James H., and Motohiro Yogo. 2005. “Testing for Weak Instruments in Linear IV Regression.” In

Identi…cation and Inference for Econometric Models: Essays in Honor of Thomas Rothenberg, edited

by Donald W.K. Andrews and James H. Stock, 80–108. Cambridge: Cambridge University Press.

Thornton, James. 2013. American Wine Economics. An Exploration of the US Wine Industry. Berkeley

and Los Angeles, California: University of California Press. Harcover.

Verhoogen, Eric. 2008. “Trade, Quality Upgrading and Wage Inequality in the Mexican Manufacturing

Sector.” Quarterly Journal of Economics 123 (2): 489–530.

Yu, Zhi. 2013. “Exchange Rate Pass-Through, Firm Heterogeneity and Product Quality: A Theoretical

Analysis.” Federal Reserve of Dallas Globalization and Monetary Policy Institute Working Paper 141.

34

Table 1: Price Breakdown for Non-EU Wine Sold in Retail Outlets in the UK

Retail price £5.76 £7.19 £8.83 £10.09

VAT (20%) £0.96 £1.20 £1.47 £1.68

Retail margin £1.92 £2.40 £2.94 £3.36

Duty £1.90 £1.90 £1.90 £1.90

Distributor margin £0.11 £0.21 £0.40 £0.51

Common Customs Tari¤ £0.11 £0.11 £0.11 £0.11

Transport £0.13 £0.13 £0.13 £0.13

Winemaker £0.63 £1.25 £1.88 £2.40

Source: Joseph (2012)

Table 2: Harmonized System Classi…cation

6-digit 12-digit Description

22.04.10 22.04.10.10.000.D Sparkling wine – Champagne variety

22.04.10.90.000.G Sparkling wine – Not Champagne variety

22.04.10.90.100.M Sparkling wine – Gassi…ed wine (i.e., aerated using CO2)

22.04.10.90.900.F Sparkling wine – Other

22.04.21 22.04.21.00.100.A Sweet wine; 2 liters

22.04.21.00.200.F Fine wine; 2 liters

22.04.21.00.900.U Other wine; 2 liters

22.04.29 22.04.29.00.100.W Sweet wine; 2 liters

22.04.29.00.200.B Fine wine; 2 liters

22.04.29.00.900.P Other wine; 2 liters

22.04.30 22.04.30.00.000.X Other grape must

Notes: The table reports the 6-digit HS codes for wine products, the corresponding 12-digit codes as used by the

Argentinean customs, and the types of wines classi…ed under each code.

Table 3: Experts Ratings

Wine Spectator (50,100) Robert Parker (50,100)

95-100 “Great” 96-100 “Extraordinary”

90-94 “Outstanding” 90-95 “Outstanding”

85-89 “Very good” 80-89 “Above average/very good”

80-84 “Good” 70-79 “Average”

75-79 “Mediocre” 60-69 “Below average”

50-74 “Not recommended” 50-59 “Unacceptable”

Notes: For both the Wine Spectator and Robert Parker rating systems, the table reports the quality bins corresponding

to the scores of the di¤erent wines.

35

Table 4: Summary Statistics on Exports Data by Year

Year Firms Wines Exports (US dollars) Observations

2002 59 794 36,504,644 2,067

2003 73 933 50,664,899 3,056

2004 107 1,171 69,640,144 3,923

2005 120 1,517 91,261,787 5,330

2006 150 1,731 112,540,681 6,793

2007 148 1,860 131,147,970 7,407

2008 148 1,804 125,851,505 6,865

2009 151 1,833 108,177,602 6,135

Total 209 6,720 725,789,235 41,576

Notes: For each year in the sample, the table reports the number of exporters, the number of wines exported, the value

of FOB exports (in US dollars), and the number of observations. The row “Total” reports the total number of …rms, the

total number of wines, the total value of FOB exports, and the total number of observations between 2002 and 2009.

Table 5: Summary Statistics

Observations Min Max Mean Median Std. dev.

Unit values (US dollars/liter) 41,576 0.02 381 5.3 3.6 6.7

Number of wines exported 41,576 1 510 139 120 130

Number of destinations 41,576 1 65 24 21 16

Wine Spectator 41,576 55 97 85 85 3.8

Parker 26,705 72 98 87 87 2.3

Notes: For each variable listed in the …rst column, the table reports its minimum, maximum, mean, median value, and

standard deviation in the sample as well as the number of observations available.

Table 6: Top Export Destinations 2002-2009

Destinations % of FOB exports

United States 30.8

Netherlands 10.3

United Kingdom 9.3

Brazil 7.2

Canada 6.4

Denmark 6.0

Finland 3.2

Sweden 3.2

Switzerland 2.8

Germany 2.3

France 1.8

Notes: The table reports total wine exports to each destination country as a

share of total wine exports over the period 2002-2009.

36

Table 7: Snapshot of the Data

Firm Year Destination Name Type Grape Vintage Quality Unit value

Wine Spectator

1 2006 United States A Red Malbec 2005 73 3.27

1 2006 United States B Red Mezcla 2003 92 13.38

1 2006 China B Red Mezcla 2003 92 11.09

Parker

2 2008 United Kingdom C Red Cabernet Sauvignon 2007 85 4.00

2 2008 United Kingdom D Red Mezcla 2004 98 43.27

2 2008 Thailand D Red Mezcla 2004 98 24.65

Notes: The …rm and wine names are con…dential and are replaced by numbers and letters instead. The table reports, for

individual wines – de…ned according to their name, grape, type, and vintage year – exported to speci…c destinations in

speci…c years, their quality rating from either the Wine Spectator or Parker and their unit value of exports (in US dollars

per liter).

Table 8: Descriptive Statistics by Quality Bin

Wine Spectator Obs. Firms Export Unit Destinations Distance GDP per

Quality Bin share value per wine capita

“Great” (95-100) 80 6 0.05 27.4 3 10,137 30,386

“Outstanding” (90-94) 5,667 68 14.81 11.9 11 9,570 26,668

“Very good” (85-89) 16,382 128 43.73 4.4 14 9,823 26,278

“Good” (80-84) 16,961 155 36.10 4.0 11 9,880 26,414

“Mediocre” (75-79) 2,282 80 4.32 3.8 6 9,655 28,012

“Not recommended” (50-74) 204 24 1.00 3.9 11 9,908 25,672

Notes: For each quality bin of the Wine Spectator, the table reports the number of observations, the number of exporters,

exports as a share of total exports (%), the average unit value (in US dollars per liter), the number of destinations per wine

exported, the average distance shipped (in kilometers), and the average income per capita of the destination countries (in

US dollars), in the sample.

Table 9: Descriptive Statistics for Exporters by Quality Bin

Firms exporting wines Obs. Firms Export Min Max Mean Wines Destinations

that are at most: share quality quality quality per wine

“Great” (95-100) 3,719 6 16.53 77 97 89 180 16

“Outstanding” (90-94) 27,919 62 67.71 60 94 85 182 13

“Very good” (85-89) 8,614 80 13.72 55 89 83 74 8

“Good” (80-84) 1,232 48 2.00 65 84 81 30 5

“Mediocre” (75-79) 91 12 0.04 72 79 77 6 3

“Not recommended” (50-74) 1 1 0.00 72 72 72 1 1

Notes: The table reports, for …rms classi…ed by Wine Spectator quality bin based on the highest quality wine the …rm

exports, the number of observations, the number of exporters, exports as a share of total exports (%), the minimum,

maximum, and mean quality exported, the number of wines, and the number of destinations per wine exported, in the

sample.

37

Table 10: Baseline Results

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Panel A: Dependent variable is unit value

ln 0140(0038)

0190(0041)

0059(0062)

– – – 0036(0064)

0190(0076)

0162(0079)

0032(0003)

– – – – – – – –

ln£ – – 0002(0000)

0002(0000)

0002(0000)

0002(0000)

0002(0000)

0002(0000)

0002(0000)

Panel B: Dependent variable is export volume

ln 1378(0516)

0868(0611)

1289(0616)

– – – 1845(0781)

0555(0695)

1780(1032)

¡0050(0006)

– – – – – – – –

ln£ – – ¡0005(0002)

¡0005(0002)

¡0004(0002)

¡0005(0002)

¡0005(0002)

¡0003(0002)

¡0002(0002)

ln 0724(0536)

0425(0633)

0420(0632)

0437(0642)

– – 1031(0793)

¡0193(0720)

1202(1057)

ln 2167(0265)

2315(0335)

2294(0334)

1943(0332)

– – 2178(0356)

1442(0499)

1328(0567)

Fixed E¤ects

Firm-dest Yes Yes Yes Yes – Yes Yes Yes Yes

Firm-year Yes – – – – – – – –

Product-year – Yes Yes Yes Yes Yes Yes Yes Yes

Firm-dest-year – – – – Yes – – – –

Dest-year – – – – – Yes – – –

Grape Yes – – – – – – – –

Type Yes – – – – – – – –

Vintage year Yes – – – – – – – –

Province Yes – – – – – – – –

HS Yes – – – – – – – –

Sample Full Full Full Full Full Full 2003/09 2002/07 2003/07

Observations 41,576 41,576 41,576 41,576 41,576 41,576 39,509 28,576 26,509

Notes: In (4), the real exchange rate is interacted with year dummies (not reported). Robust standard errors are adjusted

for clustering at the product-level in parentheses. , , and indicate signi…cance at the 1, 5, and 10 percent levels,

respectively. Unit values are in pesos per liter and export volumes are in liters. The quality ratings are from the Wine

Spectator.

38

Table 11: Income per Capita

(1) (2) (3) (4) (5)

Panel A: Dependent variable is unit value

ln 1631(0548)

2139(0594)

1155(0606)

1815(0557)

1975(0671)

ln 0126(0152)

0140(0152)

¡0071(0180)

0073(0156)

¡0103(0186)

ln£ ¡0011(0006)

¡0011(0006)

¡0010(0006)

¡0013(0006)

¡0012(0006)

ln£ ln ¡0173(0060)

¡0156(0061)

¡0144(0062)

¡0188(0062)

¡0145(0064)

ln £ ¡0001(0002)

¡0001(0001)

¡0001(0002)

0000(0002)

¡0001(0002)

ln£ £ ln 0001(0000)

0001(0000)

0001(0001)

0002(0001)

0002(0001)

ln£ ln – ¡0077(0039)

– – ¡0084(0039)

ln£ ln – – 0028(0011)

– 0025(0011)

ln – – 0286(0105)

– 0274(0107)

Panel B: Dependent variable is export volume

ln ¡0047(2092)

¡2155(2418)

2419(2335)

¡1207(2104)

¡1223(2685)

ln 3843(0647)

3791(0648)

2681(0775)

4530(0663)

3491(0794)

ln£ 0019(0021)

0018(0021)

0019(0022)

0031(0022)

0031(0022)

ln£ ln 0111(0216)

0044(0218)

0082(0218)

0231(0218)

0135(0220)

ln £ ¡0020(0007)

¡0020(0007)

¡0021(0007)

¡0028(0007)

¡0030(0008)

ln£ £ ln ¡0002(0002)

¡0002(0002)

¡0002(0002)

¡0004(0002)

¡0004(0002)

ln 0400(0631)

0275(0643)

0710(0639)

0238(0645)

0372(0666)

ln£ ln – 0304(0161)

– – 0328(0159)

ln£ ln – – ¡0050(0042)

– ¡0049(0043)

ln – – 1366(0451)

– 1233(0457)

Sample Full Full Full Excl. US Excl. US

Observations 41,576 41,576 41,180 36,714 36,318

Notes: Firm-destination and product-year …xed e¤ects are included. Robust standard errors are adjusted for clustering

at the product-level in parentheses. , , and indicate signi…cance at the 1, 5, and 10 percent levels, respectively. Unit

values are in pesos per liter and export volumes are in liters. The quality ratings are from the Wine Spectator.

39

Table 12: High versus Low Income Countries

(1) (2) (3) (4) (5)

Panel A: Dependent variable is unit value

ln£ 0303(0124)

1083(0330)

¡0219(0254)

0317(0124)

0667(0405)

ln£ 0010(0068)

0891(0342)

¡0357(0189)

0052(0073)

0646(0392)

ln£ £ ¡0000(0001)

¡0000(0001)

0000(0001)

0000(0001)

0000(0001)

ln£ £ 0002(0001)

0002(0001)

0002(0001)

0002(0001)

0002(0001)

ln£ ln – ¡0102(0038)

– – ¡0105(0039)

ln£ ln – – 0033(0011)

– 0030(0011)

ln – – 0188(0069)

– 0197(0071)

Panel B: Dependent variable is export volume

ln£ 1003(0722)

¡0971(1571)

3468(1226)

0988(0704)

1151(1892)

ln£ 1384(0628)

¡0849(1660)

3573(1016)

1438(0650)

1009(1899)

ln£ £ 0000(0003)

0000(0003)

0000(0003)

¡0001(0003)

¡0001(0003)

ln£ £ ¡0006(0002)

¡0006(0002)

¡0006(0002)

¡0008(0002)

¡0008(0002)

ln 0416(0635)

0283(0654)

0703(0647)

0259(0650)

0353(0683)

ln 2283(0337)

2305(0337)

0979(0532)

2297(0345)

1130(0540)

ln£ ln – 0243(0162)

– – 0272(0160)

ln£ ln – – ¡0048(0045)

– ¡0049(0045)

ln – – 1433(0457)

– 1323(0461)

Sample Full Full Full Excl. US Excl. US

Observations 41,576 41,576 41,180 36,714 36,318

Notes: Firm-destination and product-year …xed e¤ects are included. Robust standard errors are adjusted for clustering

at the product-level in parentheses. , , and indicate signi…cance at the 1, 5, and 10 percent levels, respectively. Unit

values are in pesos per liter and export volumes are in liters. The quality ratings are from the Wine Spectator.

40

Table 13: Endogeneity

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Panel A: Dependent variable is unit value

ln 0157(0040)

0158(0040)

0159(0040)

0026(0062)

¡0998(0561)

¡0985(0562)

0111(0066)

¡1048(0491)

¡1048(0491)

0033(0003)

0022(0010)

0022(0010)

0033(0003)

0021(0010)

0022(0010)

– – –

ln£ – – – 0002(0000)

0014(0007)

0014(0007)

0001(0000)

0015(0006)

0015(0006)

Panel B: Dependent variable is export volume

ln 1131(0528)

1180(0553)

1180(0553)

1456(0535)

¡0915(1848)

¡0907(1847)

0982(0632)

¡1085(2046)

¡1086(2046)

¡0049(0006)

¡0100(0040)

¡0100(0040)

¡0050(0007)

¡0074(0039)

¡0072(0039)

– – –

ln£ – – – ¡0004(0002)

0025(0021)

0025(0021)

¡0004(0002)

0022(0024)

0021(0024)

ln 0516(0549)

0562(0578)

0561(0578)

0511(0548)

0579(0582)

0578(0582)

0287(0643)

0333(0611)

0333(0611)

ln 2599(0275)

2609(0263)

2609(0263)

2573(0274)

2772(0304)

2770(0303)

2629(0344)

2795(0333)

2795(0333)

Estimator OLS IV IV OLS IV IV OLS IV IV

CPI O¢cial O¢cial Cavallo O¢cial O¢cial Cavallo O¢cial O¢cial Cavallo

Observations 37,723 37,723 37,723 37,723 37,723 37,723 37,723 37,723 37,723

Notes: Firm-destination, …rm-year, grape, type, vintage year, province, and HS …xed e¤ects are included in (1)-(6). Firm-

destination and product-year …xed e¤ects are included in (7)-(9). Robust standard errors are adjusted for clustering at the

product-level in parentheses. , , and indicate signi…cance at the 1, 5, and 10 percent levels, respectively. Unit values

are in pesos per liter and export volumes are in liters. The quality ratings are from the Wine Spectator. Instruments

include the monthly average temperatures and total rainfall per province over the growing period (September to March)

and the altitude of each province.

Table 14: Nonlinearities

(1) (2) (3) (4)

Dependent variable Unit value Unit value Export volume Export volume

ln 0042(0101)

0067(0062)

1230(0694)

1290(0616)

ln£ (50¡ 74) 0002(0001)

– ¡0004(0005)

–

ln£ (75¡ 79) 0002(0001)

– ¡0004(0005)

–

ln£ (80¡ 84) 0002(0001)

– ¡0004(0004)

–

ln£ (85¡ 89) 0002(0001)

– ¡0004(0004)

–

ln£ (90¡ 94) 0002(0001)

– ¡0005(0004)

–

ln£ (50¡ 94) – 0001(0000)

– ¡0005(0002)

ln£ (95¡ 100) 0006(0003)

0006(0003)

¡0002(0004)

¡0003(0003)

ln – – 0424(0631)

0417(0632)

ln – – 2308(0334)

2293(0334)

Observations 41,576 41,576 41,576 41,576

Notes: Firm-destination and product-year …xed e¤ects are included. Robust standard errors are adjusted for clustering

at the product-level in parentheses. , , and indicate signi…cance at the 1, 5, and 10 percent levels, respectively. Unit

values are in pesos per liter and export volumes are in liters. The quality ratings are from the Wine Spectator.

41

Table 15: Distribution Costs and Productivity

(1) (2) (3) (4)

Panel A: Dependent variable is unit value

ln 0102(0081)

¡0101(0168)

¡0013(0072)

0063(0116)

ln£ 0001(0000)

0003(0001)

0002(0001)

0001(0001)

Panel B: Dependent variable is export volume

ln ¡0821(1187)

3111(1247)

0208(0718)

2921(1282)

ln£ ¡0003(0003)

¡0003(0006)

¡0007(0002)

¡0002(0003)

ln ¡0930(1201)

1270(1299)

¡0830(0769)

2181(1262)

ln 4184(0658)

1003(0499)

2674(0414)

1494(0559)

Importers’ wage High Low All All

Exporters’ size All All Large Small

Observations 20,708 20,472 20,819 20,757

Notes: Firm-destination and product-year …xed e¤ects are included. Robust standard errors are adjusted for clustering

at the product-level in parentheses. , , and indicate signi…cance at the 1, 5, and 10 percent levels, respectively. Unit

values are in pesos per liter and export volumes are in liters. The quality ratings are from the Wine Spectator. Columns

(1) and (2) report the results for the high and low wage destination countries, while columns (3) and (4) report the results

for the large and small exporters, respectively.

Table 16: The Quality of Exporters and Export Market Shares

(1) (2) (3) (4)

Panel A: Dependent variable is unit value

ln ¡0002(0102)

0132(0075)

0047(0065)

¡0064(0155)

ln£ 0003(0001)

¡0000(0001)

0002(0000)

0002(0001)

Panel B: Dependent variable is export volume

ln 1267(0898)

1432(0851)

1257(0799)

0569(1694)

ln£ ¡0005(0003)

¡0006(0003)

¡0005(0002)

¡0001(0003)

ln 0160(0916)

0755(0878)

0571(0823)

¡0349(1766)

ln 2498(0481)

2079(0465)

2263(0454)

2763(0799)

Exporters’ quality High Low All All

Exporters’ market shares All All High Low

Observations 20,721 20,855 22,898 18,678

Notes: Firm-destination and product-year …xed e¤ects are included. Robust standard errors are adjusted for clustering

at the product-level in parentheses. , , and indicate signi…cance at the 1, 5, and 10 percent levels, respectively. Unit

values are in pesos per liter and export volumes are in liters. The quality ratings are from the Wine Spectator. Columns

(1) and (2) report the results for the high and low quality exporters, while columns (3) and (4) report the results for the

high and low market share exporters, respectively.

42

Table 17: The Measurement of Quality and Unrated Wines

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Panel A: Dependent variable is unit value

ln 0168(0043)

¡0000(0115)

¡0253(0189)

0101(0067)

0153(0033)

0148(0033)

0046(0057)

0166(0035)

0167(0035)

ln£ 0006(0003)

0002(0001)

0005(0002)

0002(0000)

– – 0002(0000)

0008(0002)

0006(0002)

– – – – ¡0009(0001)

¡0012(0002)

– – –

ln£ – – – – ¡0000(0000)

¡0000(0000)

– – –

Panel B: Dependent variable is export volume

ln 0941(0608)

3421(0813)

2547(1508)

1325(0631)

2116(0546)

2223(0545)

2948(0555)

2377(0535)

2131(0546)

ln£ ¡0022(0009)

¡0017(0005)

¡0020(0007)

¡0006(0002)

– – ¡0008(0002)

¡0029(0009)

¡0005(0008)

– – – – 0069(0013)

0095(0017)

– – –

ln£ – – – – 0000(0000)

0000(0000)

– – –

ln 0422(0631)

1350(0765)

0337(1361)

0256(0646)

1690(0559)

1769(0559)

1728(0548)

1731(0548)

1583(0562)

ln 2301(0334)

2114(0420)

2580(0686)

2310(0343)

1923(0276)

2017(0275)

1891(0271)

1893(0271)

1976(0280)

Sample Full Full Same bin Excl. US Full Full Mean Mean Unrated

Quality WS [1,6] Parker WS WS Rank Rank WS WS [1,6] WS [1,6]

Observations 41,576 26,705 8,982 36,714 86,882 86,882 67,589 67,589 64,302

Notes: Firm-destination and product-year …xed e¤ects are included. Robust standard errors are adjusted for clustering at the

product-level in parentheses. , , and indicate signi…cance at the 1, 5, and 10 percent levels, respectively. Unit values are in

pesos per liter and export volumes are in liters.

Table 18: The Nature of Wine

(1) (2) (3) (4)

Dependent variable Unit value Unit value Export volume Export volume

ln 0025(0067)

0077(0062)

1957(0657)

1041(0616)

ln£ 0002(0001)

0001(0000)

¡0005(0003)

¡0004(0002)

ln£ – ¡0004(0003)

– 0061(0010)

ln – – 1167(0636)

0454(0631)

ln – – 1802(0319)

2312(0332)

Sample No vintage Full No vintage Full

Observations 44,147 41,576 44,147 41,576

Notes: Firm-destination and product-year …xed e¤ects are included. Robust standard errors are adjusted for clustering

at the product-level in parentheses. , , and indicate signi…cance at the 1, 5, and 10 percent levels, respectively. Unit

values are in pesos per liter and export volumes are in liters. The quality ratings are from the Wine Spectator.

43

Table 19: Exchange Rate

(1) (2) (3)

Panel A: Dependent variable is unit value

ln 0117(0070)

0059(0062)

0026(0065)

ln£ 0001(0000)

0002(0000)

0002(0000)

ln – – 0277(0062)

Panel B: Dependent variable is export volume

ln 1570(0685)

1289(0616)

0827(0613)

ln£ ¡0007(0002)

¡0005(0002)

¡0005(0002)

ln 0525(0693)

0420(0632)

0349(0630)

ln 2717(0375)

2294(0334)

1798(0341)

ln – – 2031(0655)

Sample Excl. US dollar peg Full Full

Exchange Rate Real Real Nominal

CPI O¢cial Cavallo O¢cial

Observations 34,372 41,576 41,576

Notes: Firm-destination and product-year …xed e¤ects are included. Robust standard errors are adjusted for clustering

at the product-level in parentheses. , , and indicate signi…cance at the 1, 5, and 10 percent levels, respectively. Unit

values are in pesos per liter and export volumes are in liters. The quality ratings are from the Wine Spectator.

Table 20: Robustness

(1) (2) (3) (4) (5) (6)

Panel A: Dependent variable is unit value

ln 0076(0061)

0014(0054)

0058(0062)

0019(0066)

0059(0062)

0041(0061)

ln£ 0002(0000)

0001(0000)

0001(0000)

0001(0000)

0001(0000)

0002(0000)

Panel B: Dependent variable is export volume

ln 0953(0603)

0918(0328)

1382(0632)

1203(0660)

1296(0616)

–

ln£ ¡0005(0002)

¡0003(0002)

¡0005(0002)

¡0005(0002)

¡0005(0002)

–

ln 0037(0621)

0334(0344)

¡0922(0893)

¡0940(0886)

0426(0632)

–

ln 2316(0332)

2150(0363)

2702(0353)

2669(0371)

2291(0335)

–

Sample Intensive Quarterly Full Full All …rms Full

Lags None None One Two None None

Unit values Pesos Pesos Pesos Pesos Pesos US dollars

Observations 35,594 58,016 41,576 41,576 41,632 41,576

Notes: Firm-destination and product-year …xed e¤ects are included. In (2), the product-year …xed e¤ects are replaced by

product-quarter dummy variables. Robust standard errors are adjusted for clustering at the product-level in parentheses., , and indicate signi…cance at the 1, 5, and 10 percent levels, respectively. The quality ratings are from the Wine

Spectator.

44

Figure 1: Argentina’s Total Wine Exports (million US dollars)

Sources: Nosis and United Nations Comtrade

Figure 2: Comparison between the Wine Spectator and Parker ratings

45

Figure 3: Argentinean peso per US dollar, January 2002 to December 2009

Source: International Financial Statistics

46

Appendix Table A1: First-Stage Instrumental Variables Regressions

(1) (2) (3) (4)

Regression in Table 13 Col. (4) Panel A Col. (4) Panel A Col. (4) Panel B Col. (4) Panel B

Dependent Variable ln£ ln£

Temperature September ¡0111(0095)

¡0687(0207)

¡0110(0096)

¡0658(0207)

Temperature October ¡0177(0050)

0038(0119)

¡0177(0050)

0020(0119)

Temperature November 0358(0146)

¡0217(0323)

0358(0146)

¡0198(0323)

Temperature December ¡0189(0089)

¡0326(0215)

¡0190(0090)

¡0314(0215)

Temperature January 0287(0087)

¡0222(0180)

0287(0087)

¡0216(0180)

Temperature February 0183(0065)

0038(0172)

0183(0065)

0042(0172)

Temperature March ¡0579(0077)

¡0334(0166)

¡0578(0077)

¡0336(0166)

Rainfall September ¡0022(0004)

¡0008(0010)

¡0022(0004)

¡0008(0010)

Rainfall October 0014(0002)

¡0008(0006)

0014(0002)

¡0008(0006)

Rainfall November 0003(0002)

¡0002(0006)

0003(0002)

¡0002(0006)

Rainfall December 0004(0003)

¡0017(0006)

0004(0003)

¡0017(0006)

Rainfall January ¡0004(0001)

¡0003(0001)

¡0004(0001)

¡0003(0001)

Rainfall February ¡0004(0001)

0001(0003)

¡0004(0001)

0001(0003)

Rainfall March ¡0002(0001)

0006(0004)

¡0002(0001)

0005(0004)

Temperature September£ ln ¡0001(0017)

0202(0142)

¡0001(0017)

0206(0142)

Temperature October£ ln ¡0017(0010)

¡0135(0055)

¡0017(0010)

¡0133(0056)

Temperature November£ ln 0005(0024)

0275(0179)

0005(0024)

0280(0179)

Temperature December£ ln ¡0007(0017)

0071(0101)

¡0007(0017)

0055(0102)

Temperature January£ ln ¡0003(0020)

0042(0121)

¡0003(0020)

0032(0122)

Temperature February£ ln 0009(0019)

0083(0136)

0008(0019)

0092(0136)

Temperature March£ ln 0009(0021)

¡0160(0129)

0008(0021)

¡0151(0129)

Rainfall September£ ln ¡0002(0001)

¡0008(0010)

¡0002(0001)

¡0008(0010)

Rainfall October£ ln 0000(0000)

0003(0004)

0000(0000)

0004(0004)

Rainfall November£ ln 0001(0001)

0003(0004)

0001(0001)

0003(0004)

Rainfall December£ ln ¡0002(0000)

0003(0003)

¡0002(0000)

0003(0003)

Rainfall January£ ln ¡0001(0000)

¡0003(0001)

¡0000(0000)

¡0003(0001)

Rainfall February£ ln 0000(0000)

¡0006(0003)

0000(0001)

¡0006(0003)

Rainfall March£ ln ¡0000(0000)

¡0003(0003)

¡0000(0000)

¡0003(0003)

Altitude£ ln 0000(0000)

0003(0001)

0000(0000)

0003(0001)

ln 0016(1155)

72351(7657)

0775(1424)

71017(8170)

ln – – 0800(0832)

¡2604(2259)

ln – – 0131(0445)

¡7303(1030)

Notes: Firm-destination, …rm-year, grape, type, vintage year, province, and HS …xed e¤ects are included. Robust standard

errors are adjusted for clustering at the product-level in parentheses. , , and indicate signi…cance at the 1, 5, and 10

percent levels, respectively.

47